Cognitive Wireless Networks Using The CSS Technology

Lecture Notes in Electrical Engineering 384
MeilingLi
AnhongWang
Jeng-ShyangPan
Cognitive
Wireless
Networks
Using the CSS
Technology
Lecture Notes in Electrical Engineering

Volume 384
Board of Series editors

Leopoldo Angrisani, Napoli, Italy
Marco Arteaga, Coyoacn, Mxico
Samarjit Chakraborty, Mnchen, Germany
Jiming Chen, Hangzhou, P.R. China
Tan Kay Chen, Singapore, Singapore
Rdiger Dillmann, Karlsruhe, Germany
Haibin Duan, Beijing, China
Gianluigi Ferrari, Parma, Italy
Manuel Ferre, Madrid, Spain
Sandra Hirche, Mnchen, Germany
Faryar Jabbari, Irvine, USA
Janusz Kacprzyk, Warsaw, Poland
Alaa Khamis, New Cairo City, Egypt
Torsten Kroeger, Stanford, USA
Tan Cher Ming, Singapore, Singapore
Wolfgang Minker, Ulm, Germany
Pradeep Misra, Dayton, USA
Sebastian Mller, Berlin, Germany
Subhas Mukhopadyay, Palmerston, New Zealand
Cun-Zheng Ning, Tempe, USA
Toyoaki Nishida, Sakyo-ku, Japan
Bijaya Ketan Panigrahi, New Delhi, India
Federica Pascucci, Roma, Italy
Tariq Samad, Minneapolis, USA
Gan Woon Seng, Nanyang Avenue, Singapore
Germano Veiga, Porto, Portugal
Haitao Wu, Beijing, China
Junjie James Zhang, Charlotte, USA
About this Series

Lecture Notes in Electrical Engineering (LNEE) is a book series which reports
the latest research and developments in Electrical Engineering, namely:
Communication, Networks, and Information Theory

Computer Engineering
Signal, Image, Speech and Information Processing
Circuits and Systems
Bioengineering
LNEE publishes authored monographs and contributed volumes which present

cutting edge research information as well as new perspectives on classical elds,
while maintaining Springers high standards of academic excellence. Also
considered for publication are lecture materials, proceedings, and other related
materials of exceptionally high quality and interest. The subject matter should be
original and timely, reporting the latest research and developments in all areas of
electrical engineering.
The audience for the books in LNEE consists of advanced level students,
researchers, and industry professionals working at the forefront of their elds. Much
like Springers other Lecture Notes series, LNEE will be distributed through
Springers print and electronic publishing channels.
More information about this series at http://www.springer.com/series/7818
Meiling Li Anhong Wang

Jeng-Shyang Pan
Cognitive Wireless Networks

Using the CSS Technology
123
Meiling Li
Taiyuan University of Science
and Technology
Taiyuan
Peoples Republic of China
Jeng-Shyang Pan
Fujian University of Technology
Fuzhou, Fujian
Anhong Wang
Taiyuan University of Science
and Technology
Taiyuan
ISSN 1876-1100
ISSN 1876-1119 (electronic)
Lecture Notes in Electrical Engineering
ISBN 978-3-319-31094-7
ISBN 978-3-319-31095-4 (eBook)
DOI 10.1007/978-3-319-31095-4
Library of Congress Control Number: 2016934678
Springer International Publishing Switzerland 2016
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microlms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specic statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.
Printed on acid-free paper
This Springer imprint is published by Springer Nature
The registered company is Springer International Publishing AG Switzerland
Preface
With the rapid development of wireless networks, more and more radio spectrum
resources will be needed. Cognitive radio (CR) is an exciting emerging technology
to improve spectrum efciency, by which the licensed spectrum resources can be
shared dynamically by cognitive users. The accurate and effective spectrum sensing
technologies are key to realizing the cognitive radio, which are still research
hot-spots in the wireless sphere.
The aim of this book is to provide some useful methods to improve the spectrum
sensing performance in a systematic way, and point out an effective method for the
application of cognitive radio technology in wireless communications. After giving
a state-of-the-art survey, we propose some new cooperative spectrum sensing
(CSS) methods, with an attempt to achieve better performance. For each CSS, the
main idea and their corresponding algorithm design are elaborated in detail.
This book covers the fundamental concepts and the core technologies of CSS,
especially its latest developments. Each chapter is presented in a self-sufcient and
independent way so that the reader can select the chapters interesting to them. The
methodologies are described in detail so that the readers can repeat the corresponding experiments easily.
For researchers, it would be a good book to understand the classications of
CSS, inspiring new ideas about the novel CSS technology for CR, and a quick way
to learn new ideas from the current status of CSS. For engineers, it would be a good
guidebook to develop practical applications for CSS.
Chapter 1 provides a broad view of CR. Chapter 2 shows the CSS technologies
and current researches. Chapter 3 focuses on the CSS based on hard combination,
mainly devoted to the relationship of each performance parameter. Chapters 46 are
devoted to algorithms to solve the actual existing problems, mainly focusing on the
current research fruits of the authors. We provide the basic frameworks and the
experimental results, which may help the readers nd some new ideas. Chapter 7
introduces the application of CR and provides a basic realization method for mobile
communications.
vi
Preface
This work was supported in part by the National Natural Science Foundation of
China (No. 61272262), National Science Foundation for Young Scientists of
Shanxi Province, China (Grant No. 2014021021-2) and Doctor Startup Foundation
of TYUST, China (No. 20122032).
We are grateful to the Springer in-house editors for the editorial assistance and
excellent cooperative collaboration to produce this important scientic work. We
hope that the reader will share our excitement to present this book and will nd it
useful.
Taiyuan, Shanxi
November 2015
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . .
1.1 Dissertation Background . . . . . . . . . . .
1.2 Cognitive Radio Technology . . . . . . . .
1.2.1 Spectrum Sensing . . . . . . . . . .
1.2.2 Spectrum Management . . . . . . .
1.2.3 Spectrum Mobility. . . . . . . . . .
1.2.4 Spectrum Sharing . . . . . . . . . .
1.2.5 Conclusions . . . . . . . . . . . . . .
1.3 Spectrum Sensing Technology. . . . . . .
1.3.1 Spectrum Sensing Classication
1.3.2 Spectrum Sensing Method . . . .
1.4 Motivation . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
3
4
5
6
7
8
10
10
12
15
16
2 CSS Technology. . . . . . . . . . . . . . . . . . .
2.1 Introduction . . . . . . . . . . . . . . . . . .
2.2 Cooperative Communication Model . .
2.2.1 Cooperative Diversity . . . . . .
2.2.2 Relay Cooperation . . . . . . . . .
2.3 CSS Model . . . . . . . . . . . . . . . . . . .
2.3.1 Cooperative Diversity for CSS
2.3.2 Relay Diversity for CSS. . . . .
2.4 The Process of CSS . . . . . . . . . . . . .
2.5 Research Status . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
23
23
24
24
25
30
30
32
34
37
39
3 The Relationship Among the Performance Parameters in CSS . . . . .

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
43
45
.
.
.
.
.
.
.
.
.
.
.
vii
viii
Contents
3.3 The CSS Detection Performance . . . . . . . . . . . . . .

3.3.1 Local Detection Performance . . . . . . . . . . .
3.3.2 CSS Performance Based on Decision Fusion.
3.3.3 Simulation Results. . . . . . . . . . . . . . . . . . .
3.4 The CSS Secondary Throughput . . . . . . . . . . . . . .
3.4.1 Spectrum Utilization . . . . . . . . . . . . . . . . .
3.4.2 Secondary Throughput . . . . . . . . . . . . . . . .
3.4.3 The Optimal Algorithm . . . . . . . . . . . . . . .
3.4.4 Simulation Results. . . . . . . . . . . . . . . . . . .
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
46
46
48
50
53
53
56
58
60
66
66
Censoring Based CSS . . . . . . . . .

Introduction . . . . . . . . . . . . . . . .
System Model . . . . . . . . . . . . . . .
The C-CSS Detection Performance
4.3.1 Performance Analysis . . . .
4.3.2 Optimal Algorithm . . . . . .
4.3.3 Simulation Results. . . . . . .
4.4 The C-CSS Secondary Throughput
4.4.1 Performance Analysis . . . .
4.4.2 Optimal Algorithm . . . . . .
4.5 Summary . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
69
69
71
72
72
76
79
83
83
84
86
92
92
5 CSS Technology with Relay . . . . . . . . .

5.1 Introduction . . . . . . . . . . . . . . . . .
5.2 System Model . . . . . . . . . . . . . . . .
5.2.1 Problem Description . . . . . .
5.2.2 Signal Model . . . . . . . . . . .
5.3 Proposed Best Relay CSS Scheme . .
5.3.2 The SINR-BRCS Scheme . . .
5.3.3 Proposed Pe_BRCS Scheme .
5.3.4 Complexity Analysis . . . . . .
5.3.5 Detection Performance . . . . .
5.3.6 Simulation Results. . . . . . . .
5.4 Proposed C-BR-CSS Scheme . . . . .
5.4.2 System Model. . . . . . . . . . .
5.4.3 Performance Analysis . . . . .
5.4.4 Simulation Results. . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
95
95
97
97
99
101
101
103
103
105
105
106
108
108
109
111
116
4 The
4.1
4.2
4.3
Contents
5.5 Proposed Adaptive CSS Scheme

5.5.1 Problem Description . . .
5.5.2 Algorithm Design . . . . .
5.5.3 Simulation Results. . . . .
5.6 Summary . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . .
ix
with Best Relay

............
............
............
............
............
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
118
118
119
120
122
122
Based on Soft Combination. . . . . . . . . . . . . . . . . . .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
SC-EF-CSS Performance . . . . . . . . . . . . . . . . . . . . .
6.3.1 Performance Analysis . . . . . . . . . . . . . . . . . .
6.3.2 The Optimal Algorithm Based on N-P Criterion
6.3.3 The Optimal Algorithm Based on MDC . . . . . .
6.3.4 Simulation Results. . . . . . . . . . . . . . . . . . . . .
6.4 SC-DF-CSS Performance . . . . . . . . . . . . . . . . . . . . .
6.4.1 Performance Analysis . . . . . . . . . . . . . . . . . .
6.4.2 The Optimal Algorithm Based on N-P Criterion
6.4.3 The Optimal Algorithm Based on MDC . . . . . .
6.4.4 Simulation Results. . . . . . . . . . . . . . . . . . . . .
6.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
125
125
126
127
127
131
133
134
136
136
138
140
141
143
143
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
145
145
147
147
148
150
150
153
155
155
158
158
6 CSS
6.1
6.2
6.3
7 The SS Application in ICIC . . . . . . . .

7.1 Introduction . . . . . . . . . . . . . . . .
7.2 Problem Analysis. . . . . . . . . . . . .
7.2.1 The SFR Scheme . . . . . . .
7.2.2 Interference Analysis . . . . .
7.3 The ICI Coordination Based on SS
7.3.1 Scheme Design . . . . . . . . .
7.3.2 Interference Analysis . . . . .
7.3.3 Detection Algorithm . . . . .
7.4 Summary . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Abbreviations
3GPP
AF
AWGN
BPSK
BRCS
CC
C-CSS
CR
CRC
CRN
CSS
DF
DF
DSA
E3
EF
EGC
erfc
FC
FCC
IEEE
LRT
LTE
MAC
MRC
NP
OFDM
OFDMA
PU
SC
3rd Generation Partnership Project

Amplify and Forward
Additive White Gaussion Noise
Binary Phase Shift Keying
Best Relay Cooperation Scheme
Coded Cooperation
Censoring based CSS
Cognitive Radio
Cyclic Redundancy Check
Cognitive Radio Networks
Cooperative Spectrum Sensing
Decode and Forward
Decision Fusion
Dynamic Spectrum Access
End-to-End Efciency
Energy Fusion
Equal Gain Combining
The Complementary Error Function
Fusion Center
Federal Communications Commission
Institute of Electrical and Electronics Engineers
Likelihood Ratio Test
Long-Term Evolution
Media Access Control
Maximum Ratio Combining
NeymanPearson
Orthogonal Frequency Division Multiplexing
Orthogonal Frequency Division Multiple Access
Primary User
Soft Combine
xi
xii
SDC
SFR
SINR
SNR
SON
SR
SU
TDMA
WiMAX
xG
Abbreviations
Selective Diversity Combining

Soft Frequency Reuse
Signal to Interference and Noise Ratio
Signal-to-Noise Ratio
Self-organization Network
Secondary Relay
Secondary User
Time Division Multiple Access
Worldwide Interoperability for Microwave Access
Next Generation
Chapter 1
Introduction
In this chapter, the background of the cognitive radio technology is stated. The key
technologies are then introduced simply. Finally, the classication and methods of
the spectrum sensing technology are described.
1.1
Dissertation Background
With the rapid development of wireless networks, the demand of radio spectrum
resources also grow with each passing day, which results that the scarcity of
wireless spectrum resource is also becoming increasingly prominent. In the current
spectrum allocation framework, there are a lot of frequency bands being long-term
idle in time and space [17], which results the lower spectrum utilization. Visibly,
the spectrum scarcity problem is not only the actual lack of physical resources, but
also being related to the current wireless spectrum allocation mode. In the existing
xed spectrum allocation mode, each frequency band is xedly assigned to the
different authorized institutions. Other unlicensed users can not use those idle
spectrum resources, even if the licensed spectrum resources are not being used
temporarily. For this reason, this part of the idle spectrum resources can not be fully
utilized which severely limited the utilization of spectrum resources and restricted
the development of wireless communication. Therefore, how to enhance the utilization of the radio spectrum is one of the hot issues in the elds of domestic and
international communication at home and abroad. Dynamic spectrum access
(DSA) technology can effectively alleviate the contradiction between the low
spectrum utilization and the scarcity of spectrum resources, in which the unlicensed
users can use the free spectrum resources in time or space when assuring not
bringing influences to the licensed users communications, so as to achieve the
reuse of the spectrum resources [814]. Cognitive radio (CR) is a kind of frequency
reuse technologies, which can achieve the dynamic spectrum sharing and improve
the spectrum efciency by DSA [1520]. The concept of CR originates from the
M. Li et al., Cognitive Wireless Networks Using the CSS Technology,
Lecture Notes in Electrical Engineering 384,
DOI 10.1007/978-3-319-31095-4_1
Introduction
Pioneering research of Dr. Joseph Mitola III [21]. It has been quickly get wide
attentions from all around the world after the concept of CR was proposed.
Thereafter the spectrum management departments, the standardization organizations and the research institutions all over the world have launched their researches
on CR [2227].
The standards related to CR include IEEE 802.22, 802.16 h and IEEE P1900.
The IEEE 802.22 working group is the rst standardization organization based on the
CR technology in the world [28], which is mainly to plan the air interface standard for
the cognitive users when accessing the broadcasting bands and assure that the
broadcasting services can not be interfered [27]. The IEEE 802.22 standard is of great
signicance to the development of CR technology. The IEEE 802.16 working group
has committed to the research of wireless broadband access technology, i.e. WiMAX
(Worldwide Interoperability for Microwave Access), which is limited to the spectrum
resources. For this reason, the 802.16 h working group was established, the purpose
of which is that the series of 802.16 standards can be applied in the license-free band
by using CR technology [29]. In order to make a further study of CR, IEEE established IEEE P1900 standard group [3032], which committed to the next generation
wireless communication technology and advanced spectrum management technology. This working group has great signicance to the development of CR technology
and the co-ordination and co-existence with other wireless communication systems.
The future mobile communication network will be a ubiquitous and heterogeneous
network mode, with the ability of self-conguration, self-optimization and
self-learning in the future [33]. Recently, in order to accommodate the need of the
future networks development, the network should perform the self-organization
functions, such as self-conguration or self optimization in IEEE 802.16m. The
concept and requirement of the self-organizing network (SON) have also been
proposed in LTE-Advanced. It requires that the network has the ability of
self-conguration, self-optimization or self-healing [33], which mainly rely on the
CR technology [34, 35]. At present, the discussion of SON is still in the preliminary
stage in the IEEE and LTE organization. There are still a lot of related technologies
which need in-depth study so that the SON can be applied in actual system.
Not only the standardization organizations have recognized the enormous
potential of CR, but also the academic communities have done widely researches to
the CR technology. Professor Simon Haykin, an international famous scholar,
dened the concept of CR from the perspective of signal processing [36, 37]. He
pointed out that the CR is an intelligent wireless communication system, which can
be aware of its surrounding environment (i.e., its outside world), and uses the
methodology of understanding-by-building to learn from the environment and adapt
its internal states to statistical variations in the incoming radio frequency
(RF) stimuli by making corresponding changes in certain operating parameters
(e.g., transmit power, carrier frequency, and modulation strategy) in real time, with
two primary objectives in mind: (1) Highly reliable communications whenever and
wherever needed; and (2) Efcient utilization of the radio spectrum. The denitions
have been put forwarded by Simon Haykin promote a lot of extensive studies on
CR technology in the eld such as academic world, industries, research institutions
1.1 Dissertation Background
and industry. A lot of projects have also been started, such as the CORVUS system
jointly developed by the University of California, Berkeley and the Technology
University of Berlin, Germany [38], the OCRA systems developed by the
Polytechnic University of Georgia, USA [19], the XG (next Generation) project and
E3 (End to End Efciency) project [39] that developed by U.S. military DARPA
[40, 41]. Under the support of these projects, the CR has obtained some achievements in the eld of basic theory, network architecture, and its protocol design in
wireless communication systems.
Compared with foreign, the research on CR technology has also received highly
attention in our country. The national 973 program, the national 863 program and
the National Natural Science Foundation of China all set up a special issue of CR.
Tsinghua University, Beijing University of Post and Telecommunications,
University of Electronic Science and Technology, Hong Kong University Science
& Technology and other universities have also make in-depth studies on CR
technology. ZTE, Huawei and other companies has also been involved in the
related work in the standards development of IEEE 802.22. Judging from the
momentum of the development at home and abroad, the study on CR technology is
still in the ascendant. CR technology, as the next big thing in the wireless
communication area, opens up an effective way to solve the low utilization of the
spectrum resources and the shortage of spectrum resource. However, the study of
CR is still at a preliminary stage, there are still a lot of challenges to deal with
before the nal application such as the system architecture, protocol architecture,
standards, the specic key techniques etc.
1.2
Cognitive Radio Technology
CR technologies provide the capability to use or share the spectrum in an opportunistic manner. DSA technologies allow the CR to operate in the best available
channel. The term, CR, can formally be dened as follows [2, 19]:
A Cognitive Radio is a radio that can change its transmitter parameters based
on interaction with the environment in which it operates. From this denition, two
main characteristics of the cognitive radio can be dened [19, 36, 42]:
Cognitive capability: Cognitive capability refers to the ability of the radio
technology to capture or sense the information from its radio environment. This
capability cannot simply be realized by monitoring the power in some frequency
band of interest. More sophisticated techniques are required in order to capture
the temporal and spatial variations in the radio environment and avoid interference to other users, the vacant spectrum at a specic time or location can be
identied through this capability. Consequently, the best spectrum and appropriate operating parameters can be selected [19].
Recongurability: The cognitive capability provides spectrum awareness
whereas recongurability enables the radio to be dynamically programmed
Introduction
according to the radio environment. More specically, the CR can be programmed to transmit and receive on a variety of frequencies and to use different
transmission access technologies supported by its hardware design [19, 43].
The ultimate objective of the CR is to obtain the best available spectrum through
cognitive capability and recongurability as described before. Since most of the
spectrum has already been assigned, the most important challenge is to share the
licensed spectrum without interfering with the transmission of other licensed users.
The CR enables the usage of temporally unused spectrum, which is referred to as
spectrum hole or white space [36]. If this band is further used by a licensed user, the
CR moves to another spectrum hole or stays in the same band, altering its transmission power level or modulation scheme to avoid interference. In the CR system,
there are two kinds of users: the primary user (PU) and the cognitive user. The PU
belongs to the authorized system that has the absolute priority of possession to the
authorized spectrum resources. The cognitive users also known as the secondary
users (SUs) can access the idle spectrum opportunistically to realize communication.
In summary, the main functions for CR can be summarized as follows:
(1) Determine which portions of the spectrum is available and detect the presence
of licensed users when a user operates in a licensed band (spectrum sensing);
(2) Select the best available spectrum to meet user communication requirements
(spectrum management);
(3) Coordinate access to this channel with other users (spectrum sharing);
(4) Vacate the channel when a licensed user is detected (spectrum mobility).
1.2.1
Spectrum Sensing
Cognitive capability relies to a large extent on the awareness of the environment

dynamics. For dynamic spectrum sharing, important information is related to the
spectrum used by other users in the network, which should not be interfered when
they are primary users. Alternatively, they can be users of equal regulatory status
that occupy the channel. These other users are not necessarily equipped with
cognitive capability, and more important, they can be much more powerful. In all
cases, it is necessary to become aware of the presence of the different other users.
When communicating, the SU will cause harmful interference to receivers in its
interference range (Fig. 1.1). As a result, the sensing node should be able to detect
whether receivers are present in this range. Alternatively, when the SU is mainly
interested in avoiding interference by other users in the network, it has to determine
if there are harmful transmitters in the neighborhood. If the SU is currently not in a
transmitter detection area (Fig. 1.1), it can communicate without interference [44].
1.2 Cognitive Radio Technology
Primary base station
Secondary user
Primary user
Transmitter detection area
Receiver detection area

=
Interference area
Fig. 1.1 Detection area with and without interference
Spectrum sensing has been identied as a key enabling functionality to ensure

that SUs would not interfere with PUs, by reliably detecting PU signals. In addition,
reliable sensing plays a critical role on communication links of SUs since it creates
spectrum opportunities for them. In order to efciently utilize the available
opportunities, SUs must sense frequently all degrees of freedom (time, frequency,
and space) while minimizing the time spent in sensing [4550]. We will introduce
the spectrum sensing technology for detail in the following subsection.
1.2.2
Spectrum Management
After spectrum sensing, the spectrum management is needed to capture the best
available spectrum to satisfy user communication requirements. The spectrum
management functions mainly include spectrum analysis and spectrum decision.
In cognitive radio networks (CRN), the available spectrum holes show different
characteristics which vary over time. Since the SUs are equipped with the CR based
physical layer, it is important to understand the characteristics of different spectrum
bands. Spectrum analysis enables the characterization of different spectrum bands,
which can be exploited to get the spectrum band appropriate to the user requirements [19].
Once all available spectrum bands are characterized, appropriate operating
spectrum band should be selected for the current transmission considering the QoS
requirements and the spectrum characteristics. The spectrum decision is about
whether and how to access the spectrum. The goal of the spectrum decision is to
best meet the user communication requirements, while satisfying a set of constraints, e.g., the acceptable interference which can be created to other users in the
spectrum. The optimization goal, i.e., the outcome that best meets the user
requirements, can be a local or a global criterion. Next, it is important that the
spectrum decision is coordinated through SUs in the network. The classication of
the spectrum decision is shown in Fig. 1.2. The spectrum decision can impact the
future spectrum sensing and hence the amount of learning the wireless communication scene [44].
Introduction
Spectrum decision
Optimization behaviour
Cooperative
Noncooperative
Architecture
Distributed
Coordination
Centralized
Common
channel
Control only
No common
channel
Data channel
Fig. 1.2 Spectrum decision classication
1.2.3
Spectrum Mobility
In CRN, spectrum mobility arises when current channel conditions become worse
or a PU appears. Spectrum mobility gives rise to a new type of handoff in CRN that
we refer to as spectrum hand off. The protocols for different layers of the network
stack must adapt to the channel parameters of the operating frequency. Moreover,
they should be transparent to the spectrum hand off and the associated latency. As
pointed out in earlier sections, a CR can adapt to the frequency of operation.
Therefore, each time a SU changes its frequency of operation, the network protocols are going to shift from one mode of operation to another. The purpose of
spectrum mobility management in CRN is to make sure that such transitions are
made smoothly and as soon as possible such that the applications running on a SU
perceive minimum performance degradation during a spectrum hand off. It is
essential for the mobility management protocols to learn in advance about the
duration of a spectrum hand off. This information should be provided by the sensing
algorithm. Once the mobility management protocols learn about this latency, their
job is to make sure that the ongoing communications of a SU undergo only minimum performance degradation. Consequently, multi-layer mobility management
protocols are required to accomplish the spectrum mobility functionalities. These
protocols support mobility management adaptive to different types of applications.
For example, a TCP connection can be put to a wait state until the spectrum hand
off is over. Moreover, since the TCP parameters may change after a spectrum hand
off, it is essential to learn the new parameters and ensure that the transitions from
the old parameters to new parameters are carried out rapidly. For a data communication e.g., FTP, the mobility management protocols should implement mechanisms to store the packets that are transmitted during a spectrum hand off, whereas
Dynamic spectrum
access
Dynamic licensing
(Dynamic exclusive use)
Dynamic sharing
(coexistence)
Horizontal sharing
Heterogeneous
networks
Symmetric
Homogeneous
networks
Vertical sharing
Underlay
Overlay
Asymmetric
Fig. 1.3 Dynamic spectrum access, classication along regulatory status
for a real-time application there is no need to store the packets as the stored packets,
if delivered later, will be stale packets and can not be used by the corresponding
application [19].
1.2.4
Spectrum Sharing
As introduced in the previous section, DSA technology is needed to achieve a better

use of the spectrum. DSA is the opposite of the current static spectrum management
policy. However, various approaches are possible to make the spectrum management more adaptive, as presented in Fig. 1.3 [44].
Dynamic licensing results in a dynamic spectrum allocation that gives exclusive
use to the technology or network that currently has the most prot of spectrum use.
It is similar to the current spectrum regulation in that it licenses spectrum bands for
exclusive use. This dynamic licensing is, however, much more flexible, to be able
to adapt to the wireless communication dynamics [44].
Ideally, spectrum sharing should adapt very fast to all dynamics present in
wireless communication, which can be caused by the channel variations or because
of the burst application demands. Coexistence or dynamic sharing allows such
sharing, in theory, on a packet per packet basis since it licenses spectrum to networks simultaneously, while relying on in-network spectrum sharing techniques to
avoid conflicts. This model for spectrum sharing assumes that all networking nodes
have equal regulatory status. As a result, this model is also referred to as open
Introduction
sharing model [51] or as spectrum commons [52, 53]. Medium access protocols for
wireless networks are working according to this model and considerable literature
can be found on both centrally controlled or distributed access techniques for
spectrum sharing between nodes of a single network [44].
In the above-discussed case for coexistence, both considered networks adapt
their transmission schemes as function of the environment. This is because both
networks can benet from avoiding the mutual interference, and both networks
have (limited) adaptive or cognitive capabilities. In this case, spectrum sharing can
be classied as symmetric in nature. This asymmetric spectrum sharing, in which
only one of the technologies is adaptive, is somewhat similar to vertical spectrum
sharing [44].
The initial denition of spectrum sharing assumes the existence of a PU and a
SU. While the spectrum has been licensed to the PU only, the SU can use it
opportunistically provided this does not affect the PUs performance. Two
approaches exist for spectrum access to minimize the interference caused to the PUs
by the SUs communication: spectrum overlay and spectrum underlay. Overlay
spectrum sharing refers to the spectrum access technique used. More specically, a
node accesses the network using a portion of the spectrum that has not been used by
licensed users [19, 5460]. As a result, interference to the primary system is
minimized. Underlay spectrum sharing exploits the spread spectrum techniques
developed for cellular networks [61]. Once a spectrum allocation map has been
acquired, a SU begins transmission such that its transmit power at a certain portion
of the spectrum is regarded as noise by the licensed users. This technique requires
sophisticated spread spectrum techniques and can utilize increased bandwidth
compared to overlay techniques.
1.2.5
Conclusions
The core idea of CR technology is that the SUs can intelligently detect and analyze
the spectrum in wireless environment, and then identify the free spectrum in
specic time or specic position. By this way, the SUs can access the selected
optimal spectrum to realize spectrum sharing and the improved spectrum utilization
[6264]. To achieve this goal, the SUs have to implement spectrum sensing to nd
the PU and detect the idle spectrum resources. Spectrum analysis is needed after the
completion of spectrum sensing. The SUs will analyze the characteristics of the
available spectrum resources in the CRN so that they can determine the appropriate
frequency bands for radio transmission. The key of spectrum analysis is how to
select the appropriate spectrum according to the results of the spectrum sensing to
realize reliable communication. The purpose of the spectrum decision is making
preparations for spectrum sharing [19, 65] by conrming the access parameters
(including modulation encoding and the transmitted power) based on the spectrum
Fig. 1.4 Cognitive ring
sensing and the spectrum analysis. Spectrum analysis and spectrum decision
belongs to the scope of spectrum management. Once the work band is selected, the
SU can transmit the signal in this band, however, the SU access the spectrum in an
opportunistic way in CRN, because the available spectrum are changed dynamically
in different space and time [6671]. When the PU appears, the SU must release the
occupied spectrum and switch to other available spectrum to achieve seamless hand
off, which is called spectrum hand off, named as spectrum shift. In the process of
spectrum hand off, how to ensure the continuity and reliability of the SUs transmission and realize the fast seamless hand off are difcult, which rstly require that
the SU should continuously perform spectrum sensing to detect the availability of
the free spectrum resources. Multiple SUs may use the same idle spectrum to
communicate simultaneously when the idle spectrum is detected, therefore, in order
to avoid the SUs interference with each other, spectrum sharing is needed. In the
spectrum sharing, the SUs transmissions can be appropriately coordinated and
managed by certain resource allocation strategies which are similar to the MAC
media access protocols in the existing systems [7173]. The above process usually
can be described by a cognitive ring, as shown in Fig. 1.4. In short, the goal of CR
is to provide available spectrum resources for SUs when assuring sufcient protection to the PU. In order to achieve this goal, various aspects of the cognitive ring
need to be considered, in which the spectrum sensing is the premise and the key of
the CR technology. It is very important to study the spectrum sensing technology so
as to make full use of the limited spectrum resources.
In summary, the SUs need to search spectrum constantly until the idle spectrum
are detected in the process of spectrum sensing. When the PU returns to the channel
which has been occupied by a SU, the SU have to release this channel to avoid
interference to the PU. However, it is hard to guarantee that the PU signals can be
detected absolutely up to 100 % in actual. Therefore, the core problem of spectrum
sensing technology is how to detect the idle spectrum accurately.
10
1.3
1.3.1
Introduction
Spectrum Sensing Technology

Spectrum Sensing Classication
The spectrum sensing is to detect the unused spectrum bands in order to enable CR
to adapt its environment. Since there exist multiple SUs in CRN, according to
whether they implement spectrum sensing collaboratively, the spectrum sensing
schemes can be categorized as cooperative spectrum sensing scheme and
non-cooperative spectrum sensing scheme. Further, cooperative spectrum sensing
can be divided into centralized sensing [7476], distributed sensing [77] and
relay-assisted sensing [7880] on how cooperative CR users share the sensing data
in the network, as shown in Fig. 1.5 [81].
In centralized cooperative spectrum sensing, a central identity called fusion
center (FC) controls the three-step process of cooperative spectrum sensing. Firstly,
the FC selects a channel or a frequency band of interest for sensing and instructs all
cooperating SUs to individually perform local sensing. Secondly, all cooperating
SUs report their sensing results via the control channel. Then the FC combines the
received local sensing information, determines the presence of PUs, and transmits
the decision back to cooperating SUs. As shown in Fig. 1.6a, SU0 is the FC and
SU1SU5 are cooperating SUs performing local sensing and reporting the results to
SU0. For local sensing, all SUs are tuned to the selected licensed channel or
frequency band where a physical point-to-point link between the PU transmitter and
each cooperating SU for observing the primary signal is called a sensing channel.
For data reporting, all SUs are tuned to a control channel where a physical
point-to-point link between each cooperating SU and the FC for sending the sensing
results is called a reporting channel. Note that centralized cooperative spectrum
sensing can occur in either centralized or distributed CRN. In centralized CRN, a
Spectrum sensing
Cooperative
sensing
Centalized
sensing
Distributed
sensing
Fig. 1.5 Spectrum sensing classication
Non-Cooperative
sensing
Relay-assisted
sensing
1.3 Spectrum Sensing Technology
(a)
11
(b)
(c)
PU
PU
PU
SU0(FC)
SU0(FC)
SU5
SU5
SU5
SU1
SU1
SU1
SU4
SU2
SU3
SU4
SU4
SU2
SU3
SU2
(Relay)
SU3
(Relay)
Fig. 1.6 Classication of cooperative spectrum sensing
CR base station (BS) is naturally the FC. Alternatively, in CR ad hoc networks

where a CR BS is not present, any SU can act as a FC to coordinate cooperative
spectrum sensing and combine the sensing information from the cooperating
neighbors [81].
Unlike centralized cooperative spectrum sensing, distributed cooperative spectrum sensing does not rely on a FC for making the cooperative decision. In this
case, SUs communicate among themselves and converge to a unied decision on
the presence or absence of PUs by iterations. Figure 1.6b illustrates the cooperation
in the distributed manner. After local sensing, SU1SU5 share the local sensing
results with other users within their transmission range. Based on a distributed
algorithm, each SU sends its own sensing data to other users, combines its data with
the received sensing data, and decides whether or not the PU is present by using a
local criterion. If the criterion is not satised, SUs send their combined results to
other users again and repeat this process until the algorithm is converged and a
decision is obtained. In this manner, this distributed scheme may take several
iterations to reach the unanimous cooperative decision [81].
In addition to centralized and distributed cooperative spectrum sensing, the third
scheme is relay-assisted cooperative spectrum sensing. Since both sensing channel
and reporting channel are not perfect, a SU who experience a weak sensing channel
and a strong reporting channel and a SU who experience a strong sensing channel
and a weak reporting channel, for example, can complement and cooperate with
each other to improve the performance of cooperative spectrum sensing. In
Fig. 1.6c, SU1, SU4, and SU5, who observe strong PU signals, may suffer from a
weak report channel. SU2 and SU3, who have a strong reporting channel, can serve
as relays to assist in forwarding the sensing results from SU1, SU4, and SU5 to the
FC. In this case, the reporting channels from SU2 and SU3 to the FC can also be
called relay channels. Note that although Fig. 1.6c shows a centralized structure, the
relay-assisted cooperative spectrum sensing can exist in distributed scheme. In fact,
when the sensing results need to be forwarded by multiple hops to reach the
destination, all the intermediate hops are relays. Thus, if both centralized and
distributed structures are one-hop cooperative spectrum sensing, the relay-assisted
structure can be considered as multi-hop cooperative spectrum sensing. In addition,
12
Introduction
Spectrum sensing
Transmitter
Energy detection
Matched filter
Receiver
Interference
temperature
Cyclostationary
Detection
Fig. 1.7 Spectrum sensing method
the relay for cooperative spectrum sensing here serves a different purpose from the
relays in cooperative communications [71], where the CR relays are used for forwarding the PU trafc [81].
1.3.2
Spectrum Sensing Method
In cooperative spectrum sensing, each SU can utilize the sensing method in

non-cooperative spectrum sensing to implement the spectrum sensing, the detail
classication is shown as in Fig. 1.7. The current non-cooperative spectrum sensing
methods include: the sensing based on PU receiver, the sensing based on interference temperature model and the sensing based on the PU transmitter [19]. The
main idea of the sensing based on the interference temperature model is that in a
certain frequency band, if all of the energy that the PU received from the interference source do not exceed a predened maximum limit (also known as interference temperature limit), the SU can share this band with the PU. The interference
temperature model manages interference at the receiver through the interference
temperature limit, which is represented by the amount of new interference that the
receiver could tolerate. In other words, the interference temperature model accounts
for the cumulative RF energy from multiple transmissions and sets a maximum
threshold on their aggregate level. As long as SUs do not exceed this limit by their
transmissions, they can use this spectrum band. The difculty of this detection
model lies in how to effectively measure the interference temperature. A SU is
naturally aware of its transmit power level and its precise location with the help of a
positioning system. With this ability, however, its transmission could cause signicant interference at a neighboring receiver on the same frequency. However,
currently, there exists no practical way for a cognitive radio to measure or estimate
13
the interference temperature on nearby primary receivers. Since primary receivers

are usually passive devices, a SU cannot be aware of the precise locations of
primary receivers. Therefore, if SUs cannot effectively measure their transmission
on all possible receivers, a useful interference temperature measurement may not be
feasible [19]. The sensing based on PU receiver is also known as local oscillator
leakage detection, in which, whether there exist available spectrum resources are
judged by detecting the target users who are receiving signal in the authorization
system, it mainly makes a judgment by using local oscillation leakage energy that
released by RF front-end when the PUs receiving equipment are working, those
leaked signal energy are often very weak, so this method is only applied to test
television receiver at present. In practice, it is difcult for SU to obtain accurate
position information of the PU receiver [50, 8287]. The sensing based on the PU
transmitter is that the SU determine whether there are idle spectrum by detecting the
transmitted signal of the PU. This method of spectrum sensing is simple and easy to
operate. Therefore, most existing spectrum sensing algorithms focus on the
detection of the primary transmitted signal based on the local observations of the
CR. The typical detection methods based on primary transmitted signal include:
energy detection, matched lter detection and cyclostationary detection [19, 71].
1.3.2.1
Energy Detection
If prior knowledge of the PU signal is unknown, the energy detection method is

optimal for detecting any zero-mean constellation signals [88]. In the energy
detection approach, the radio-frequency (RF) energy in the channel or the received
signal strength indicator is measured to determine whether the channel is idle or not.
First, the input signal is ltered with a band-pass lter to select the bandwidth of
interest. The output signal is then squared and integrated over the observation
interval. Lastly, the output of the integrator is compared to a predened threshold to
infer the presence or not of the PU signal, as shown in Fig. 1.8. When the spectrum
is analyzed in the digital domain, fast Fourier transform (FFT) based methods are
used. Specically, the received signal xt, sampled in a time window, is rst
passed through an FFT device to get the power spectrum jX f j2 . The peak of the
power spectrum is then located. After windowing the peak of the spectrum, we get
jY f j2 . The signal energy is then collected in the frequency domain. Although the
energy-detection approach can be implemented without any prior knowledge of the
PU signal, it still has some drawbacks. The rst problem is that it has poor performance under low SNR conditions. This is because the noise variance is not
accurately known at the low SNR, and the noise uncertainty may render the energy
detection useless [88]. Another challenging issue is the inability to differentiate the
Fig. 1.8 Energy detection

model
14
Introduction
interference from other secondary users who share the same channel with the PU
[87]. Furthermore, the threshold used in energy selection depends on the noise
variance, and small noise power estimation errors can result in signicant performance loss [71]. As energy detection method is simple and easy to realize, it has
been widely applied in practice.
1.3.2.2
Matched Filter Detection
When a SU has a prior knowledge of the PU signal, the optimal signal detection is a
matched lter, as it maximizes the signal-to-noise ratio (SNR) of the received
signal. A matched lter is obtained by correlating a known signal, or template, with
an unknown signal to detect the presence of the template in the unknown signal.
This is equivalent to convolving the unknown signal with a time-reversed version
of the template [71]. So the matched lter can be considered as an optimal linear
lter that maximizes the SNR in the presence of additive noise [50, 89, 90].
The main advantage of matched lter is that it needs less time to achieve high
processing gain due to coherent detection [88]. Another signicant disadvantage of
the matched lter is that it would require a dedicated sensing receiver for all
primary user signal types. In the CR scenario, however, the use of the matched lter
can be severely limited since the information of the PU signal is hardly available at
the CRs. The use of this approach is still possible if we have partial information of
the PU signal such as pilot symbols or preambles, which can be used for coherent
detection [15]. For instance, to detect the presence of a digital television
(DTV) signal, we may detect its pilot tone by passing the DTV signal through a
delay-and-multiply circuit. If the squared magnitude of the output signal is larger
than a threshold, the presence of the DTV signal can be detected [71].
1.3.2.3
Cyclostationary Detection
Cyclostationary detection is more robust to noise uncertainty than energy detection.

If the signal of the PU exhibits strong cyclostationary properties, it can be detected
at very low SNR values by exploiting the information (cyclostationary feature)
embedded in the received signal. A signal is said to be cyclostationary (in the wide
sense) if its autocorrelation is a periodic function of time t with some period [71,
91]. The cyclostationary detection can be performed as follows.
First, the cyclic auto-correlation function (CAF) of the observed signal xt is
calculated as E xt sx t sej2pat , where Efg denotes the statistical
expectation operation and a is called the cyclic frequency.
The spectral correlation function (SCF) Sf ; a is then obtained from the discrete
Fourier transformation of the CAF. The SCF is also called cyclic spectrum,
which is a two-dimension function in terms of frequency f and cyclic frequency a.
15
The detection is completed by searching for the unique cyclic frequency corresponding to the peak in the SCF plane.
This detection approach is robust to random noise and interference from other
modulated signals because the noise has only a peak of SCF at the zero cyclic
frequency and the different modulated signals have different unique cyclic frequencies. In [92], the cyclostationary detection method is employed for the
detection of the Advanced Television Systems Committee DTV signals in wireless
region-area network systems. Experimental results show superior detection performance even in very low SNR region. In [93], distributed detection is considered
for scanning spectrum holes, where each CR employs a generalized likelihood ratio
test for detecting primary transmissions with multiple cyclic frequencies.
The above approach can detect the PU signal from other SUs signals over the
same frequency band provided that the cyclic features of the PU and the CR signals
differ from each other, which is usually the case, because different wireless systems
usually employ different signal structures and parameters. By exploiting the distinct
cyclostationary characteristics of the PU and the CR signals, a strategy of extracting
channel-allocation information is proposed in spectrum pooling systems [94], where
the PU is a GSM network and the CR is an OFDM-based WLAN system. However,
cyclostationary detection is more complex to implement than the energy detection
and requires a prior knowledge of PU signal such as modulation format [71].
1.4
Motivation
Since there exists the hidden terminal problem in CRN as shown in Fig. 1.9, when
the channel between the PU and the SU is effected by the shadow of building, the
SU cannot accurately detect whether the PU is present or not in the interested bands
of the SU. As a result, the SU may incorrectly detect the PU signal whereas the
spectrum is idle, which results that the SU can not use this frequency band and thus
lose the opportunities to access it. On the other hand, when the band is occupied by
the PU while the PU signal is not detected by the SU, the SU will access this band
which causes serious interference to the PU. Accordingly, when the channel
between the PU and the SU exists multipath fading and shadowing, the PU signal
Fig. 1.9 The hidden terminal

problem between the PU and
the SU
PU
SU
16
Introduction
can not be accurately detected by single-user spectrum sensing. To address this

issue, multiple CRs can be designed to collaborate in spectrum sensing [15, 71].
Cooperative communication technology can effectively resist wireless channels
fading and greatly improve the transmission reliabilities, which has been regarded
as one of the ways to deal with the problems of CR technology [95, 96]. References
[22, 23, 78, 79] show that the system performance can be improved by introducing
the cooperative communication into the CR system, in which, both of the advantages can be utilized. The cooperative spectrum sensing (CSS) is a main application
in the combination of CR technology and cooperative communication technology.
Previous studies have shown that the CSS can effectively improve the detection
performance under fading channels [15, 24, 25, 75, 80, 97104]. As discussed
above, from the perspective of the network architecture, the CSS can be divided
into centralized based CSS and distributed based CSS. In the centralized CSS mode,
there exist a control channel and a central controller which is also called the FC, in
which, each SU transmit the sensing result to the FC by the control channel, and
then the FC makes a nal decision on whether the PU signal exists or not by
collecting all of the results from each SU and broadcast the availability of detected
spectrum in the whole network. In the distributed CSS mode, there is no central
controller, each SU sharing its local sensing information and decide whether the PU
signal is present by combining its local sensing information with others. In the
distributed CSS mode, each SU will perform information fusion, which results a
large amount of data computing. From the perspective of spectrum sharing, the FC
is responsible for allocating the idle spectrum resources in the centralized CSS;
while in the distributed CSS, each node is self-centered and mainly satisfy the need
of itself, it will access the spectrum as long as the idle spectrum is detected without
considering about the needs of other SUs. We mainly concentrate on the centralized based CSS in this book.
References
1. Federal Communications commission (2002) Spectrum policy task force, ET Docket no.
02-135 memorandum opinion and order
2. Federal Communications commission (2003) Notice of proposed rule making and order, ET
Docket No 03-222
3. Federal Communications commission (2003.) Notice of inquiry and notice of proposed
Rulemaking, ET Docket No. 03-237
4. Cheng P, Deng RL, Chen JM (2012) Energy-efcient cooperative spectrum sensing in
sensor-aided cognitive radio networks. IEEE Wirel Commun 19(6):100105
5. McHenry M, Livsics E, Nguyen T et al (2007) XG dynamic spectrum access eld test results.
IEEE Commun Mag 45(6):5157
6. McHenry MA NSF (2005) Spectrum occupancy measurements project summary. Shared
Spectrum Company Rep
7. Deng RL, Chen JM, Yuen C et al (2012) Energy-efcient cooperative spectrum sensing by
optimal scheduling in sensor-aided cognitive radio networks. IEEE Trans Veh Technol 61
(2):716725
References
17
8. Buddhikot MM, Ryan K (2005) Spectrum management in coordinated dynamic spectrum

access based cellular networks. In: Proceedings of IEEE DySPAN 2005, pp 299307
9. Etkin R, Parekh A, Tse D (2005) Spectrum sharing for unlicensed bands. In: Proceedings of
IEEE DySPAN 2005, pp 251258
10. Grandblaise D, Bourse D, Moessner K, Leaves P (2002) Dynamic spectrum allocation
(DSA) and recongurability. In: Proceedings of software-dened radio (SDR) Forum
11. Kamakaris T, Buddhikot MM, Iyer R (2005) A case for coordinated dynamic spectrum
access in cellular networks. In: Proceedings of IEEE DySPAN 2005, pp 289298
12. Leaves P, Moessner K, Tafazoli R, Grandblaise D, Bourse D, Tonjes R, Breveglieri M
(2004) Dynamic spectrum allocation in composite recongurable wireless networks. IEEE
Comm Mag 42:7281
13. Wang P, Akyildiz IF (2011) Can dynamic spectrum access induce heavy tailed delay? In:
2011 IEEE symposium on new frontiers in dynamic spectrum access networks (DySPAN),
Aachen, pp 197207
14. Wang P, Akyildiz IF (2012) On the origins of heavy-tailed delay in dynamic spectrum access
networks. IEEE Trans Mobile Comput 11(2):204217
15. Cabric D, Mishra SM, Brodersen RW (2004) Implementation issues in spectrum sensing for
cognitive radios. In: Proceedings of 38th Asilomar conferences signals, system, computers.
United States, pp 772776
16. Cabric D, Brodersen RW (2005) Physical layer design issues unique to cognitive radio
systems. In: Proceedings of IEEE personal indoor and mobile radio communications
(PIMRC) 2005
17. Cabric D, Mishra SM, Willkomm D, Brodersen R, Wolisz A (2005) A cognitive radio
approach for usage of virtual unlicensed spectrum. In: Proceedings of 14th IST mobile and
wireless communications summit
18. Maldonado D, Lie B, Hugine A, Rondeau TW, Bostian CW (2005) Cognitive radio
applications to dynamic spectrum allocation. In: Proceedings of IEEE DySPAN 2005,
pp 597600
19. Akyildiz IF, Lee WY, Vuran MC et al (2006) Next generation/dynamic spectrum
access/cognitive radio wireless networks: a survey. Comput Net J 50(13):21272159
20. Shin KG, Hyoil K, Min AW et al (2010) Cognitive radios for dynamic spectrum access: from
concept to reality. IEEE Wirel Commun 17(6):6474
21. Mitola J, Maguit QG (1999) Cognitive radio: making software radios more personal. IEEE
Pers Commun 6(4):1318
22. Simeone O, Gambini J, Bar-Ness Y et al (2007) Cooperation and cognitive radio. In:
IEEE international conference on proceedings communications, 2007. ICC07, Glasgow,
pp 65116515
23. Scutari G, Palomar D, Barbarossa S (2008) Cognitive MIMO radio. IEEE Signal Process
Mag 25(6):4659
24. Weiss T, Jondral F (2004) Spectrum pooling: an innovative strategy for the enhancement of
spectrum efciency. IEEE Commun Mag 42(3):S8S14
25. Ganesan G, Li Y (2005) Agility improvement through cooperative diversity in cognitive
radio. In: Proceedings of IEEE global telecommunications conference. GLOBECOM05,
St. Louis, MO, United States, pp 25052509
26. Tang JM (2011) Research on transmission performance of cognitive radio system based on
cooperative communication. Beijing University of Posts and Telecommunications, Beijing
27. Yan SN (2011) Research on spectrum sharing technologies based on cooperation for
cognitive radios. Beijing University of Posts and Telecommunications, Beijing
28. IEEE 802.22-2011(TM) Standard for cognitive wireless regional area networks (RAN) for
operation in TV bands http://www.ieee802.org/22/. 1 July 2011
29. IEEE 802.16s license-exempt (LE) task group, [online]. Available: http://grouper.
ieee802.0rg/16/e
30. IEEE P1900 working group [online]. Available: http://grouper.ieee.org/groups/emc/emc/
1900/index.html
18
Introduction
31. Holland O, Muck M, Martigne P et al (2007) Development of a radio enabler for

reconguration management within the IEEE P1900.4 working group. IEEE DySPAN
2007:232239
32. Muck M, Buljore S, Martigne P et al (2007) IEEE P1900.B: coexistence support for
recongurable, heterogeneous air interfaces. IEEE DySPAN 2007:381389
33. Shen J (2004) 3GPP long term evolution: principle and system design. The Peoples Posts
and Telecommunications Press, Beijing
34. Si JB (2010) Wireless cooperative transmission and the application in cognitive radio
networks. Xidian University, Xian
35. Guo CL, Feng CY, Zeng ZM (2010) Cognitive radio network technologies and application.
Publishing House of Electronics Industry, Beijing
36. Haykin S (2005) Cognitive radio: brain-empowered wireless communications. IEEE J Sel
Areas Commun 23(2):201220
37. Zhang GW (2011) Spectrum sensing algorithms for cognitive radio networks. Shandong
University, Jinan
38. Brodersen RW et al (2004) Coruvs: a cognitive radio approach for usage of virtual
unlicensed spectrum. Berkeley Wireless Research Center (BWRC) White paper
39. FP6 End-to-End Recongurability (R2R II) Integrated Project (IP), http://www.ntia.doc.gov/
osmhome/allochrt.pdf
40. DAPRA XG WG (2003) The XG Architectural framework V1.0
41. DAPRA XG WG (2003) The XG Vision RFC V1.0
42. Thomas RW, DaSilva LA, MacKenzie AB (2005) Cognitive networks. In: Proceedings of
IEEE DySPAN 2005, pp 352360
43. Jondral FK (2005) Software-dened radio-basic and evolution to cognitive radio,
EURASIP J Wirel Commun Networking 2005(3):275283
44. Hossain E, Bhargava V (2007) Cognitive wireless communication networks. Springer, Berlin
45. Chen R, Park JM, Bian K (2008) Robust distributed spectrum sensing in cognitive radio
networks. In: Proceedings of IEEE INFOCOM, Phoenix, pp 18761884
46. Wax M, Kailath T (1985) Detection of signals by Information theoretic criteria. IEEE Trans
Acoust Speech Signal Process 33(2):387392
47. Editorial Guest (2007) Adaptive, spectrum agile and cognitive wireless networks. IEEE J Sel
48. Varshney PK (1996) Distributed detection and data fusion, 1st edn. Article Book, Springer,
Berlin pp 1276
49. Lehtomaki JJ, Vartiainen J, Juntti M, Saarnisaari H (2006) Spectrum sensing with forward
methods. In: Proceedings of IEEE MILCOM, Washington, DC, pp 17
50. Dior (2009) Spectrum sensing optimization based on detection and power constraints.
Huazhong University of Science and Technology, Wuhan
51. Zhao Q, Sadler BM (2007) Dynamic spectrum access: signal processing, networking and
regulatory policy. IEEE Signal Process Mag 55(5):22942309
52. Benkler Y (1998) Overcoming agoraphobia: building the commons of the digitally
networked environment. Harvard J Law Technol 11:(287)
53. Lehr W, Crowncroft J (2005) Managing shared access to a spectrum commons. In:
Proceedings of IEEE DySPAN 2005, pp 420444
54. Brik V, Rozner E, Banarjee S, Bahl P, DSAP (2005) A protocol for coordinated spectrum
access. In: Proceedings of IEEE DySPAN 2005, pp 611614
55. Cao L, Zheng H (2005) Distributed spectrum allocation via local bargaining. In: Proceedings
of IEEE sensor and Ad Hoc communications and networks (SECON) 2005, pp 475486
56. Ma L, Han X, Shen CC (2005) Dynamic open spectrum sharing MAC protocol for wireless
ad hoc network. In: Proceedings of IEEE DySPAN 2005, pp 203213
57. Sankaranarayanan S, Papadimitratos P, Mishra A, Hershey S (2005) A bandwidth sharing
approach to improve licensed spectrum utilization. In: Proceedings of IEEE DySPAN 2005,
pp 279288
References
19
58. Zhao J, Zheng H, Yang GH (2005) Distributed coordination in dynamic spectrum allocation
networks. In: Proceedings of IEEE DySPAN 2005, pp 259268
59. Zhao Q, Tong L, Swami A (2005) Decentralized cognitive MAC for dynamic spectrum
access. In: Proceedings of IEEE DySPAN 2005, pp 224232
60. Zheng H, Cao L (2005) Device-centric spectrum management. In: Proceedings of IEEE
DySPAN 2005, pp 5665
61. Huang J, Berry RA, Honig ML (2005) Spectrum sharing with distributed interference
compensation. In: Proceedings of IEEE DySPAN 2005, pp 8893
62. Yucek T, Arslan H (2009) A survey of spectrum sensing algorithms for cognitive radio
applications. IEEE Commun Surv Tutorials 11(1):116130
63. Wang B, Liu KJR (2011) Advances in cognitive radio networks: a survey. IEEE J Sel Top
Signal Process 5(1):523
64. Rabbachin A, Quek TQS, Hyundong S et al (2011) Cognitive network interference. IEEE J
Sel Areas Commun 29(2):480493
65. Akyildiz IF, LeeW Y, Vuran MC et al (2008) A survey on spectrum management in
cognitive radio networks. IEEE Commun Mag 46(4):4048
66. Hamid M, Mohammed A, Yang Z (2010) On spectrum sharing and dynamic spectrum
allocation: MAC layer spectrum sensing in cognitive radio networks, In: 2010 WRI
international conference on communications and mobile computing, Shenzhen, China,
pp 183187
67. Subramanian AP, Gupta H, Das S R et.al (2007) Fast spectrum allocation in coordinated
dynamic spectrum access based cellular networks. In: 2007 2nd IEEE international
symposium on new frontiers in dynamic spectrum access networks, Dublin, Ireland,
pp 320330
68. Raman C, Yates RD, Mandayam NB (2005) Scheduling variable rate links via a spectrum
server. In: 2005 1st IEEE international symposium on new frontiers in dynamic spectrum
access networks (DySPAN), Baltimore, MD, United States, pp 110118
69. Zekavat SA, Li X (2005) User-central wireless system: ultimate dynamic channel allocation.
In: 2005 1st IEEE international symposium on new frontiers in dynamic spectrum access
networks (DySPAN), Baltimore, MD, United States, pp 8287
70. Zheng H, Cao L (2005) Device-centric spectrum management. In: 2005 1st IEEE
international symposium on new frontiers in dynamic spectrum access networks
(DySPAN), Baltimore, MD, United States, pp 5665
71. Letaief KB, Zhang W (2009) Cooperative communications for cognitive radio networks.
Proc IEEE 97(5):878893
72. De Domenico A, Strinati EC, Di Benedetto M-G (2012) A survey on MAC strategies for
cognitive radio networks. IEEE Commun Surv Tutorials 14(1):2144
73. Salami G, Durowoju O, Attar A et al (2011) Comparison between the centralized and
distributed approaches for spectrum management. IEEE Commun Surv Tutorials 13
(2):274290
74. Visotsky E, Kuffner S, Peterson R (2005) On collaborative detection of tv transmissions in
support of dynamic spectrum sharing. In: Proceedings of IEEE DySPAN 2005, pp 338345
75. Amir G, Elvino SS (2005) Collaborative spectrum sensing for opportunistic access in fading
environments. In: Proceedings of 1st IEEE international symposium new front. Dynamic
spectrum networking DySPAN. Baltimore, MD, United States, pp 131136
76. Unnikrishnan J, Veeravalli VV (2008) Cooperative spectrum sensing for primary detection in
cognitive radio. IEEE J Sel Top Signal Process 2(1):1827
77. Li Z, Yu F, Huang M (2009) A cooperative spectrum sensing consensus scheme in cognitive
radios. In: Proceedings of IEEE INFOCOM 2009, pp 25462550
78. Ganesan G, Li Y (2007) Cooperative spectrum sensing in cognitive radio, Part I: two user
networks. IEEE Trans Wireless Commun 6(6):22042212
79. Ganesan G, Li Y (2007) Cooperative spectrum sensing in cognitive radio, Part II: multiuser
20
Introduction
80. Zhang W, Letaief KB (2008) Cooperative spectrum sensing with transmit and relay diversity
in cognitive radio networks-[transaction letters]. IEEE Trans Wireless Commun 7
(12):47614766
81. Akyildiz I, BrandonF L, Balakrishnan R (2011) Cooperative spectrum sensing in cognitive
radio networks: a survey. Phys Commun 4:4062
82. Chair Z, Varshney PK (1986) Optimal data fusion in multiple sensor detection systems. IEEE
Trans Aerosp Electron Syst 22(1):98101
83. Quan Z, Cui S, Sayed AH, Poor HV (2008) Wideband spectrum sensing in cognitive radio
networks. In: Proceedings of IEEE ICC, Beijing, pp 901906
84. Guo C, Zhang T, Zeng Z, Feng C (2006) Investigation on spectrum sharing technology based
on cognitive radio. In: Proceedings of rst international conference on communications and
networking in China (ChinaCom), Beijing, pp 15
85. Taherpour A, Gazor S, Kenari MN (2008) Wideband spectrum sensing in unknown white
gaussian noise. J Inst Eng Technol 2(6):736771
86. Cabric D, Tkachenko A, Broders RW (2006) Spectrum sensing measurements of pilot,
energy, and collaborative detection. In: Proceeding of IEEE MILCOM. Washington DC,
pp 17
87. Sai SN, Corderio C, Challapali K (2005) Spectrum agile radios: utilization and sensing
architectures. Proceedings of IEEE DySPAN, Baltimore, pp 160169
88. Sahai A, Hoven N, Tandra R (2004) Some fundamental limits on cognitive radio. In
Proceedings of Allerton Conferences, Monticello
89. Proakis John G (2001) Digital communications, 4th edn. McGraw-Hill, New York
90. Qihang P, Kun Z, Jun W, Shaoqian L (2006) A distributed spectrum sensing scheme based
on credibility and evidence theory in cognitive radio context. In: Proceedings of 17th annual
IEEE international symposium on personal, indoor and mobile radio communications
(PIMRC), Helsinki, pp 15
91. Gardner WA, Spooner CM (1992) Signal interception: performance advantages of
cyclic-feature detectors. IEEE Trans Commun Jan 40:149159
92. Han N, Shon S, Chung JH, Kim JM (2006) Spectral correlation based signal detection
method for spectrum sensing in IEEE 802.22 WRAN systems. In: Proceedings of
international conferences on advanced communication technology, Phoenix Park, Korea,
2022(3):17651770
93. Lunden J, Koivunen V, Huttunen A, Poor HV (2007) Spectrum sensing in cognitive radios
based on multiple cyclic frequencies. In: Proceedings of 2nd international conferences on
cognitive radio oriented wireless networking communication (CrownCom), Orlando, FL, Jul
31Aug 3
94. Oner M, Jondral FK (2007) On the extraction of the channel allocation information in
spectrum pooling systems. IEEE J Sel Areas Commun (25):558565
95. Si JB (2010) Wireless cooperative transmission and the application in cognitive radio
networks. Xidian University, Xian
96. Li MM (2011) Research on key techniques in cognitive MIMO wireless networks. Beijing
University of Posts and Telecommunications, Beijing
97. Amir G, Elvino SS (2007) Spectrum sensing in cognitive radio networks: the
cooperation-processing tradeoff. Wirel Commun Mobile Comput 7(9):10491060
98. Amir G, Elvino SS (2007) Opportunistic spectrum access in fading channels through
collaborative sensing. J Commun 2(2):7182
99. Sun CH, Zhang W, Letaief KB (2007) Cooperative spectrum sensing for cognitive radios
under bandwidth constraints. In: IEEE wireless communication networking conferences
WCNC. China, Kowloon, pp 15
100. Sun CH, Zhang W, Letaief KB (2007) Cluster-based cooperative spectrum sensing in
cognitive radio systems. In: IEEE International conference on communications, 2007.
ICC07. Glasgow, pp 25112515
101. Liang YC, Zeng YH, Peh E et al (2008) Sensing-throughput tradeoff for cognitive radio
networks. IEEE Trans Wirel Commun 7(4):13261337
References
21
102. Shen JY, Jiang T, Liu, SY et al (2009) Maximum channel throughput via cooperative
spectrum sensing in cognitive radio networks. IEEE Trans Wirel Commun 8(10): 51665175
103. Zhang W, Mallik RK, Letaief KB (2008) Cooperative spectrum sensing optimization in
cognitive radio networks. In: IEEE international conference on communications, China,
pp 34113415
104. Hillenbrand J, Weiss T, Jondral FK (2005) Calculation of detection and false alarm
probabilities in spectrum pooling systems. IEEE Commun Lett 9(4): 349351
Chapter 2
CSS Technology
In this chapter, the CSS technology is stated for detail, involving cooperative
communication model and CSS model. Specically, how to use the traditional
cooperative communication technology to solve the problems existing in cognitive
radio networks are illustrated clearly.
2.1
Introduction
Traditional wireless networks have predominantly used direct point-to-point or

point-to-multipoint (e.g., cellular) topologies. In contrast to conventional
point-to-point communications, cooperative communications allows different users
or nodes in a wireless network to share resources and to create collaboration
through distributed transmission/processing, in which each users information is
sent out not only by the user but also by the collaborating users [1, 2]. Cooperative
communications is a new communication paradigm that promises signicant
capacity and multiplexing gain increase in wireless networks [3, 4]. It also realizes a
new form of space diversity to combat the detrimental effects of severe fading [5]. It
has been recently recognized as a powerful solution that can overcome the limitation of wireless systems [6]. The basic idea behind cooperative transmission rests
on the observation that in a wireless environment, the signal transmitted or
broadcast by a source to a destination node, each employing a single antenna, is
also received by other terminals, which are often referred to as relays or partners.
The relays process and retransmit the signals they have received. The destination
node then combines the signals coming from the source and the partners, thereby
creating spatial diversity and taking advantage of the multiple receptions of the
same data at the various terminals and transmission paths. In addition, the interference among terminals can be dramatically suppressed by distributed spatial
processing technology. By allowing multiple cognitive radios to cooperate in
spectrum sensing, the hidden terminal problem can be addressed [79].
DOI 10.1007/978-3-319-31095-4_2
23
24
2 CSS Technology
Cooperative techniques have already been considered for wireless and mobile
broadband radio [10] and also have been under investigation in various IEEE 802
standards. The IEEE 802.11 standard is concerned with wireless local area networks (WLANs) in unlicensed bands in indoor environments. A recent evolution of
IEEE 802.11 using mesh networking, i.e., 802.11s is considering the update of
802.11 MAC layer operations to self-conguration and multihop topologies [11].
The IEEE 802.16 standard utilize an orthogonal frequency-division multiplexing
(OFDM), orthogonal frequency-division multiple access (OFDMA), and
single-carrier based xed wireless metropolitan-area network in licensed bands of
1066 GHz. As an amendment of 802.16 networks, IEEE 802.16j is concerned with
multihop relay to enhance coverage, throughput, and system capacity [1, 12].
However, the critical challenging issue in spectrum sensing is the hidden terminal problem, which occurs when the CR is shadowed or in severe multipath
fading. Recent work has shown that cooperative spectrum sensing (CSS) can
greatly increase the probability of detection in fading channels [1223].
2.2
2.2.1
Cooperative Communication Model

Cooperative Diversity
In [24] and [25], Sendonaris et al. introduced and examined the concept of user
cooperation diversity. The implemented strategy uses a pair of transmitting,
full-duplex users who cooperate in sending independent data from both users to a
common destination.
In cooperative wireless communication, we are concerned with a wireless network, being cellular or ad hoc variety, where the wireless agents, which we call
users, may increase their effective quality of service (measured at the physical layer
by bit error rates, block error rates, or outage probability) via cooperation. In a
cooperative communication system, each wireless user is assumed to transmit data
as well as act as a cooperative agent for another user (Fig. 2.1). Cooperation leads to
interesting trade-offs in code rates and transmit power. In the case of power, one
may argue on one hand that more power is needed because each user, when in
cooperative mode, is transmitting for both users. On the other hand, the baseline
Fig. 2.1 Cooperative

transmission model
User1
BS
User2
2.2 Cooperative Communication Model
25
transmit power for both users will be reduced because of diversity. In the face of
this trade-off, one hopes for a net reduction of transmit power, given everything else
being constant. Similar questions arise for the rate of the system. In cooperative
communication each user transmits both his/her own bits as well as some information for his/her partner; one might think this causes loss of rate in the system.
However, the spectral efciency of each user improves, because, due to cooperation
diversity the channel code rates can be increased [26].
2.2.2
Relay Cooperation
In cooperative communications, independent paths between the user and the base
station are generated via the introduction of a relay channel as illustrated in Fig. 2.2.
The relay channel can be thought of as an auxiliary channel to the direct channel
between the source and destination. A key aspect of the cooperative communication
process is the processing of the signal received from the source node done by the
relay. These different processing schemes result in different cooperative communications protocol. Cooperative communications protocols can be generally categorized into xed relaying schemes and adaptive relaying schemes. In xed
relaying, the channel resources are divided between the source and the relay in a
xed (deterministic) manner. The processing at the relay differs according to the
employed protocols. There are mainly three relaying protocols: amplify-andforward (AF), decode-and-forward (DF), and coded cooperation (CF). In AF, the
received signal is amplied and retransmitted to the destination. The advantage of
this protocol is its simplicity and low cost implementation. But the noise is also
amplied at the relay. In DF, the relay attempts to decode the received signals.
If successful, it re-encodes the information and retransmits it. In selective relaying,
Fig. 2.2 The introduction

of relay channel
Destination (BS)
Source
Relay
Relay channel
Direct channel
Source
Relay channel
Relay
Source
26
2 CSS Technology
if the signal-to-noise ratio of the signal received at the relay exceeds a certain
threshold, the relay performs decode-and-forward operation on the message. On the
other hand, if the channel between the source and the relay suffers from severe
fading such that the signal-to-noise ratio is below the threshold, the relay idles.
Moreover, if the source knows that the destination does not decode correctly, then
the source may repeat to transmit the information to the destination or the relay may
help forward information, which is termed as incremental relaying. In this case, a
feedback channel from the destination to the source and the relay is necessary [27].
In this article, we only consider a single relay helping a user (source) in the
network forwarding information. The basic steps of relay cooperation are as
follows:
Step 1 A source sends information to its destination, and the information is also
received by the relay at the same time.
Step 2 The relay can help the source by forwarding or retransmitting the
information to the destination.
2.2.2.1
AF Relaying Protocol
In a xed AF relaying protocol, the relay simply scales the received version and
transmits an amplied version of it to the destination. Each user in this method
receives a noisy version of the signal transmitted by its partner. As the name
implies, the user then amplies and retransmits this noisy version. The base station
combines the information sent by the user and partner, and makes a nal decision
on the transmitted bit (Fig. 2.3). Although noise is amplied by cooperation, the
base station receives two independently faded versions of the signal and can make
better decisions on the detection of information [27].
The AF relay channel can be modeled as follows. The signal transmitted from
the source x is received at both the relay and destination as
ys;r
Fig. 2.3 Fixed

amplify-and-forwarded relay
protocol
p
Phs;r x ns;r
2:1
Destination
Relay
Source
ys;d
27
p
Phs;d x ns;d
2:2
where, P is the transmitted power at the source, hs;r and hs;d are the channel fading
coefcients between the source and the relay and destination, respectively. The
terms ns;r and ns;d denote the additive white Gaussian noise with zero-mean and
variance N0.
In this protocol, the relay amplies the signal from the source and forwards it to
the destination ideally to equalize the effect of the channel fade between the source
and the relay. The relay does that by simply scaling the received signal by a factor
that is inversely proportional to the received power, which is denoted by [27]
p
P
br q
2
Phs;r N0
2:3
The signal transmitted from the relay is thus given by br ys;r and has power
P equal to the power of the signal transmitted from the source. To calculate the
mutual information between the source and the destination, we need to calculate the
total instantaneous SNR at the destination. The SNR received at the destination is
the sum of the SNRs from the source and relay links. The SNR from the source link
is given by [27]
2
SNRs;d Chs;d
2:4
where, C P=N0 .
In the following we calculate the received SNR from the relay link. In step 2 the
relay amplies the received signal and forwards it to the destination with transmitted
power P. The received signal at the destination in phase 2 can be described as:
p
P
yr;d q
hr;d ys;r nr;d
2
Phs;r N0
2:5
where hr;d is the channel fading coefcient from the relay to the destination and nr;d
is an additive noise. More specically, the received signal yr;d in this case is
yr;d
p
P
q
hr;d hs;r x n0r;d
2

P hs;r N0
2:6
28
2 CSS Technology
where,
n0r;d
p
P
q
hr;d ns;r nr;d
2
Phs;r N0
2:7
Assume that the noise terms ns;r and nr;d are independent, then the equivalent
noise n0r;d is a zero-mean, complex Gaussian random variable with variance
N00
!
2
Phr;d
1 N0
2
Phs;r N0
2:8
The destination receives two copies from the signal x through the source link and
relay link. There are different technologies to combine the two signals. The optimal
technology that maximizes the overall signal-to-noise ratio is the maximal ratio
combiner (MRC). Note that MRC combining requires a coherent detector that has
knowledge of all channel coefcients. The SNR at the output of the MRC is equal
to the sum of the received signal-to-noise ratios from both branches [27].
With knowledge of the channel coefcients hs;d , hs;r and hr;d , the output of the
MRC detector at the destination can be written as
y a1 ys;d a2 yr;d
2:9
It is needed to note that the combining factors a1 and a2 should be designed to

maximize the combined SNR.
2.2.2.2
DF Relaying Protocol
After the relay has received the signal from the destination, it also can process the
received signal by decoding it, re-encode it and then retransmit it to the receiver.
This kind of relaying is termed a DF relaying protocol, which is often simply called
a DF scheme without the confusion from the selective DF relaying scheme.
In this method a user attempts to detect the partners bits and then retransmits the
detected bits (Fig. 2.4). The partners may be assigned mutually by the destination. If
the decoded signal at the relay is denoted by ^x, the transmitted signal from the relay
p
can be denoted by P^x, given that ^x has unit variance. Note that the decoded signal
at the relay may be incorrect. If an incorrect signal is forwarded to the destination,
the decoding at the destination is meaningless. It is clear that for such a scheme the
diversity achieved is only one, because the performance of the system is limited by
the worst link from the source-relay and source-destination. Although xed DF
relaying has the advantage over AF relaying in reducing the effects of additive noise

Fig. 2.4 Fixed
decode-and-forwarded relay
protocol
29
Destination
Relay
Decoded bits
Source
at the relay, it entails the possibility of forwarding erroneously detected signals to

the destination, causing error propagation that can diminish the performance of the
system. The mutual information between the source and the destination is limited
by the mutual information of the weakest link between the source-relay and the
combined channel from the sourcedestination and relaydestination [27].
2.2.2.3
CF Relaying Protocol
Coded cooperation differs from the previous schemes. Coded cooperation is a

method that integrates cooperation into channel coding. Coded cooperation works
by sending different portions of each users code word via two independent fading
paths. The basic idea is that each user tries to transmit incremental redundancy to its
partner. When it is not in possible, the users automatically revert to a noncooperative mode. The key to the efciency of coded cooperation is that all of this is
managed automatically through code design, without feedback between the users
(Fig. 2.5) [26, 28, 29].
Fig. 2.5 Coded cooperation

relay protocal
Destination
Re-encoded bits
Relay
Source
30
2.3
2 CSS Technology
CSS Model
In CRN, the cooperative technologies can be used by SUs to achieve CSS so as to

improve the signal detection efciency. According to the analysis above, the CSS
model can be divided into the following two categories: (1) cooperative relay
transmission (named as relay based CSS thereafter) (2) cooperative transmission
between secondary nodes to improve spatial diversity, which further include centralized based CSS and distributed based CSS.
2.3.1
Cooperative Diversity for CSS
Multiple antennas technology has been shown as an efcient way to provide

superior reception performance due to the potential high-space diversity [30].
In CRN, it is not practical to implement multiple antennas at each SU because of the
increasing cost and hardware complexity. However, a virtual antenna array can be
formed by allowing multiple SUs to cooperate. Consequently, the classical
space-time coding approaches [31] which have been widely used in multiple-input
multiple-output (MIMO) systems can be used in CRN so as to achieve a high
cooperative diversity [9].
According to the cooperative method, there are centralized based CSS and
distributed based CSS to achieve cooperative diversity. In centralized based CSS, a
CR BS called FC controls the three-step process of CSS. First, the FC selects a
channel or a frequency band of interest for sensing and instructs all cooperating SUs
to individually perform local sensing. Second, all cooperating SUs report their
sensing results via the control channel. Then the FC combines the received local
sensing information, determines the presence of PUs, and diffuses the decision back
to cooperating SUs. While in distributed based CSS, the SUs do not rely on a FC
for making the cooperative decision and each SU act as a FC. In this case, SUs
communicate among themselves and converge to a unied decision on the presence
or absence of PUs by iterations. In this article, we only consider the centralized
based CSS as example to study the CSS performance.
Consider the case when two-located SUs cooperate on spectrum sensing, as
illustrated in Fig. 2.6. Since the two users are close, the channels between two users

diversity technique for
centralized based CSS. The
two co-located SUs exchange
their local decisions and form
a distributed antenna array
over the reporting channels
Secondary user 1
Secondary user 2
Cognitive radio
base station
2.3 CSS Model
31
can be assumed to be ideal. Firstly, the two users perform local spectrum sensing
independently and obtain the sensing results D1 and D2 for user 1 and user 2,
respectively. Then, they may exchange their decisions and send them alternatively
in two time slots, with user 1 transmitting {D1, D2} and user 2 transmitting
{D2, D1}. By doing so, each decision is reported to the common receiver through
two independent fading channels. This gives rise to a space diversity gain of 2.
When the number of SUs in cooperative spectrum sensing is K, it can be expected
that a diversity gain of K will be achieved [9].
For high data rate wireless communications, channel frequency selectivity
becomes a critical challenging issue that can signicantly affect the system performance. Orthogonal frequency division multiplexing (OFDM) is a powerful tool
that can deal with the detrimental effects of multipath fading [32] and has been
adopted in many wireless standards such as DTV and wireless LAN. In particular,
an OFDM-based cognitive radio system structure is considered by the IEEE 802.22
working group for wireless regional area networks (WRAN). For OFDM-based
cognitive radios, cooperative diversity technique can be performed as follows. The
two users exchange their local spectrum sensing decisions. Then, the decisions will
be sent through two separated sub-channels from each user to the common receiver.
By doing so, a frequency diversity gain of 2 can be achieved over frequencyselective fading channels. Therefore, by exploiting a cooperative diversity among
co-located secondary users, we can reduce the reporting error probability and then
enhance the CSS performance. However, when there are multiple SUs to detect one
PUs presence, their local spectrum sensing results have the same information about
the detected PU, in this case, the local spectrum sensing results can be directly
transmitted to the FC to achieve diversity gain without the informations exchange
by the two SUs, which have been widely studied. In this book, we also consider this
occasion of the CSS by diversity cooperation, which is also called as CSS based on
nodes cooperation (CSS for short in the following) [9].
The CSS model with N nodes is shown in Fig. 2.7. The whole process is divided
into two steps. Step (1): Each SU perform the local spectrum sensing independently
to detect the PU signal. Step (2): Each SU send the detected PU information to the
FC, and the FC combines all of the information to make a nal decision.
spectrum sensing model with
N nodes
32
2.3.2
2 CSS Technology
Relay Diversity for CSS
CR has been suggested to improve the spectrum utilization by allowing secondary

unlicensed users to opportunistically share the spectrum that is not used by the
primary licensed users. Cooperative relay technology is regarded widely as a key
technology for increasing transmission diversity gain in various types of wireless
networks, including CRNs, which can combat signal fading due to multipath
propagation in a wireless medium. The original idea of cooperative relay transmission comes from the basic relay model that consists of three terminals: a source
S, a relay R, and a destination D. By enabling a set of cooperating relays to forward
received information, the spatial diversity can be achieved through cooperation by
multiple terminals [2, 33]. In this article we investigate a simple wireless network,
where a spectrum-rich node is selected as the relay node to improve the performance between the source and the destination.
In the context of CRN, cooperative relay transmission can give rise to the
following two different but basic scenarios.
1. Cooperative relay transmission between PUs and SUs.
In this case, SUs can relay the trafc of a primary transmitter toward the intended
destination. The rationale behind such a decision is that helping PUs nish their
transmissions as quickly as possible will, in turn, lead to more transmission
opportunities for SUs [33, 34].
2. Cooperative relay transmission between SUs.
In this scenario, a SU acts as a relay from the transmission of another (source)
secondary node. When the reporting channels of some SUs experience heavy
shadowing, the local decisions in these SUs cannot be forwarded to the FC. Then,
the maximum cooperative diversity gain of CSS will be reduced. Assume that SU
i fails to send its decision to the FC due to heavy shadowing in its reporting
channel. This is the case when the received signal power is so weak that it is merged
into the noise. In this case, FC has to make a random decision if it incorporates such
an unreliable SU into the cooperative decision. Hence, using some unreliable
cognitive radios cannot improve the CSS performance. To address this issue, FC
could censor the SNR of the received signal to check whether this cognitive radio is
reliable enough before counting it into the cooperative decision. If the SNR of the
received signal from SUi is lower than a predened threshold, then the SU i will be
labeled as an unreliable one. Under the supervision of the FC, the unreliable one
can relay its local spectrum sensing result to other cognitive radios which are in
enough good channel state, as shown in Fig. 2.8 [33].
With the relay technique, we see that CSS achieves the full cooperation among
SUs by avoiding transmission of local sensing results over bad reporting channels.
Suppose that M out of K SUs experience heavy shadowing. Without any relay, the
diversity gain of the CSS is only (K-M). However, with the help of other relay
2.3 CSS Model
33
Fig. 2.8 Cooperative relay

sensing
cognitive radios, it is demonstrated that the maximum cooperative diversity gain K

can be achieved [1, 35].
We can see that cooperation transmission between SUs aims to increase the
secondary throughput for a given spectral hole, whereas cooperation transmission
between PUs and SUs aims to increase the probability of transmission opportunities. In summary, both of the cooperation transmission schemes described above try
to improve spatial diversity for the same spectrum frequency band.
In this book, we only consider the latter and do deep research on how to improve
the sensing performance by SUs cooperation. CSS is based on cognitive relay,
which is also called the relay based CSS in the following, as shown in Fig. 2.9. The
whole process is divided into three steps. Step (1): The SU detect the PU signals by
performs the local spectrum sensing. The secondary relay (SR) can choice whether
to detect the PU signals. Step (2): The SU send the detected information to the FC
and the SR. Step (3): If the SR has successfully received the information from the
SU, the SR will forward the received information to the FC. Further more, if the PU
signal has also been detected by the SR in step (1), then the SR will combine the
information received from the SU and the PUs information detected in the above
step, and then transmit the combined information to the FC. The SR can also
feedback to the SU whether it has received the correct information. This kind of
collaborative approach is actually a typical application in CR of cooperative
transmission mode. In this approach, the FC can achieve diversity gain by receiving
the two signals from the SR (including the SUs information) and the SU.
In the two CSS model described above, there exist multiple methods when the
SR and the SU forward the information to the FC, such as AF, DF and CC [2, 23].
In the CRN, the SU and the SR play not only as destination node, but also as source
node. When they are played as source nodes, they have already received the PU
signal or the SU signal and therefore they are also acted as the forwarding nodes,
spectrum sensing model based
on cognitive relay
34
2 CSS Technology
which are different between the cognitive relay cooperation communication and the
traditional relay cooperation communication. In order to describe the three data
forwarding modes described above, the nodes acted as transmission source nodes in
both models are called the source nodes and the nodes in the receive side are called
the destination nodes. For example, in the relay-based CSS, the SU is the source
node and the SR is the destination node in step (2); while in step (3), the SR is the
source node and the FC is the destination node. In the second CSS method, the SU
is the source node and the FC is the destination node. In the AF mode, the source
node simply amplies and forwards the received signals to the destination node
without any decoding or encoding operations. In the DF mode, the signal received
by the source node will be demodulated, decoded and estimated at rst, and then
transmitted to the destination node. In the CC mode, the cooperation and the
channel coding are combined, in which, the source node will decode and recode
the received signals and then transmit the recoded signal to the destination node.
The above three relay cooperative approaches have their values in different scenarios to realize respective advantages. In this article, we mainly consider the DF
mode when doing research on the relay based CSS; while during the CSS study, we
mainly consider the AF and the DF mode.
2.4
The Process of CSS
According to the description in Sect. 2.3.2, the relay-based CSS actually can be
seen as the case that the FC combines the information of two nodes. In the
relay-based CSS, the information transmitted from the SR to the FC may contain
not only the PU information but also the SU information. The overall sensing
process can be described by two consecutive processes of centralized-based CSS:
sensing process and reporting process. As shown in Fig. 2.10, the whole process
consists of three steps: local spectrum sensing, the local spectrum sensing results
transmitting and information fusion. In the following, we will take a brief
Fig. 2.10 The centralized

based CSS in CRN
2.4 The Process of CSS
35
introduction to the three steps of the CSS and point out the problem we considered
in this article.
Step 1: Local spectrum sensing
Each SU performs local single spectrum sensing utilizing certain detection
algorithms, such as energy detection algorithm, matched lter detection algorithm
and cyclostationary detection method etc. In the energy detection method, the SU
determines whether the channel is idle or not by measuring the energy of radio
frequency signal or the received signal strength indicator. Its process is simple:
rstly, the input signal is ltered with a band-pass lter to select the interesting
bandwidth. The output signal is then squared and integrated over the observation
interval. Finally, the integrator output signal is compared to a predened threshold
to infer whether the PU signal is present or not. Since the energy detection method
is simple and easy to operate, a lot of literatures have studied the local spectrum
sensing performance by using energy detection method [17, 3639]. When energy
detection method is used in the local spectrum sensing, each SU will transmit the
detected energy signal or decision results to the destination node.
Step 2: Local spectrum sensing results transmitting
In the centralized based CSS, each SU send the detected signals to the FC
through the reporting channel. Many researchers have studied the cooperative
spectrum sensing performance when the reporting channels (the channels from SUs
to the FC) are Additive White Gaussian Noise (AWGN) [1315, 20, 21, 40, 41]. As
shown in Fig. 2.10, the hidden terminal problem also exist in the reporting channels, for example, there exists shadowing between SU1 and the FC, therefore, the
data transmitted from SU1 to the FC will be impacted by the channel fading, which
may result the error transmission. We can learn from literature [1, 35] that the
fading of the reporting channel will also affect the performance of CSS. At present,
the research of CSS performance under both imperfect sensing channels and
reporting channels still at an initial stage, and we will study it deeply in this paper.
Step 3: Information fusion at the FC
In the centralized based CSS, the FC combines all of the information from each
SU and makes a nal decision to infer the presence or absence of the PU in the
observed channels. There are different ways for information fusion. The performance of a variety of fusion methods have been studied in literature [4246]. We
can conclude that the main fusion methods including soft combination and hard
combination. In the method of soft combination, the SU will be weighted before
sending information to the FC so that the channel state information can be used to
improve the accuracy of information combination; while in method of hard combination, the SU sends the information to the FC directly without any preprocessing. In the other hand, fusion method can be divided into data fusion and
decision fusion in the point of the data format having been transmitted by the SU.
36
2 CSS Technology
From Step 1 to Step 2, when each SU performs the local spectrum sensing, it can
either directly send the detected PU information to the FC or can make a judgment
rstly and then send the result to the FC, where, the former belongs to data fusion,
and the latter belongs to decision fusion.
In this paper, the SU performs local spectrum sensing using energy detection
method. If the SU directly sends its detected PU energy to the FC, we regard it as a
kind of AF method from the point of cooperative method. On the other hand, we
call it as the energy fusion (belongs to the data fusion) from the point of fusion, if
the received energy signal at the SU is directly forwarded to the FC without any
decoding or encoding operations. Each SU transmits the detected PU energy signal
to the FC, and the FC makes a nal decision by combing all of the received data
from each SU utilizing equal gain combing (EGC), maximum ratio combining
(MRC) and selection diversity combining (SDC) etc. From [43, 4750], we can see
that the better detection performance can be achieved by data fusion, however,
because the CR system is a band-limited system, if there are more SUs the more
control channel bandwidth will be occupied for sending the local sensing results,
which may result in the lower spectrum utilization.
Usually, in the CSS, a simple decision fusion method is usually used in order to
save the control channel bandwidth. Each SU will make a binary decision based on
its local observation. It indicates the presence of the PU if the local decision result is
1, and the absence of the PU is 0. Actually, this method is a kind of simple DF
method, in which the PUs signal is rstly demodulated, decoded and re-estimated
and then transmitted to the FU by the SU. In the decision fusion mode, the SU will
make a 1 bit quantization judgment according to the detected signal rstly and then
makes a nal decision according to multiple 1 bit result utilizing k out of N
criterion. Researchers have made a detailed study to the performance of k out of N
in the literature [21, 42, 51]. The or rule corresponds to the case of k = 1, in which
the FC infers the presence of the PU as long as one of the SUs assume the presence
of the PU. It can be seen that the or rule is very conservative for the SUs to access
the licensed band, which is to ensure that the interference to the PU is minimized
and to reduce the probability of miss detection. The and rule corresponds to the
case of k = n, in which the FC makes judgment that the PU is exist only when all of
the SUs infer the presence of the PU. It can be seen that the and rule is an
aggressive strategy, i.e., the SU can access the licensed spectrum as long as one of
the SUs infer the absence of the PU. Compared with the or rule the higher
spectrum efciency can be achieved in the and rule, while the probability of miss
detection also increased at the same time. The or rule gives better performance
than other rules in the point of protecting the PU. In the actual occasions of
spectrum sharing, the SU access the idle spectrum opportunistically. The PU may
return to the occupied spectrum at any time, so we must make sure that the PU is
protected sufciently. Therefore, in the research of CSS, when the decision fusion is
considered, the FC often combine the information by the or rule.
2.5 Research Status
2.5
37
Research Status
In CRN, the PU should be protected sufciently when the SUs access the spectrum.
Therefore, how to accurately sense the idle channels and detect the presence of the
PU so as to ensure the PU un-disturbed when the SUs access the channels are our
primary research objects. The hidden terminal problem is a difcult problem in
spectrum sensing [1, 20], i.e. the channel multipath fading and shadowing effects
often make the single-user spectrum sensing performance very poor, so it is very
difcult to accurately detect whether the PU signal is present or not in single-user
spectrum sensing. The detection performance under fading sensing channels can be
effectively improved by multiple SUs cooperative spectrum sensing, which has
been studied widely.
The system performance gain caused by users collaboration has been investigated in [20, 21] when the sensing channels are AWGN, in which, the two users
cooperation method and multiple-users cooperation method are given based on the
periodic spectrum sensing. Researchers have studied the affect of sensing diversity
order to the CSS performance when the sensing channel experiences AWGN
channel and fading channel respectively [1315]. The results show that the performance of CSS improves with the increase of sensing diversity. Furthermore, it
has been proved theoretically that compared with the single user spectrum sensing,
the CSS can reduce the demand of the average signal-to-noise of sensing channels.
The researches above are all based on periodic spectrum sensing, in which, the
sensing time and sensing performance are contradictory, i.e. a longer sensing
duration can produce a better sensing performance but results in a longer waiting
time for the SUs to access the channel, which will cause serious interference to the
PU [1]. Optimal sensing duration has been studied to improve the system performance in [5156]. However, the researches described above didt consider the
occasion that the reporting channels suffer from fading. In [1, 7, 35], the authors
pointed out that the reporting channels fading would deteriorate the CSS performance. In addition, in CSS, the total sensing time include two parts: the local
detection time and the reporting time for sending the SUs local information. In the
existing studies, the Time Division Multiple Access (TDMA) mode has been
considered for sending local sensing results. Obviously the fewer SUs involved in
CSS, the shorter the whole sensing duration. However, a small number of SUs in
CSS results in a small sensing diversity order which will lead to the lower probability of detection. This problem can be addressed by allowing the SUs to send the
decisions on orthogonal frequency bands, i.e. Orthogonal Frequency Division
Multiple Access (OFDMA) mode, but this requires a large portion of the available
bandwidth [1], consequently the lower frequency utilization, which should be
studied further.
In CSS, more SUs participated in cooperation will make the higher sensing
diversity order and the better sensing performance. In the decision based CSS, the
control bandwidth can be greatly reduced by one bit quantization compared with
data fusion and multiple bits quantization method. However, when the number of
38
2 CSS Technology
sensing users is very large, the total number of sensing bits transmitted to the FC is
still very huge and the larger control bandwidth will be occupied. Consequently,
how to save the control bandwidth as soon as possible and improve the spectrum
utilization at the same time deserved further research. In addition, in [35], it has
been shown that the influence of reporting channels fading to the sensing performance is also related to the sensing diversity order. Under certain reporting
channels fading, the higher sensing diversity order will result the worse sensing
performance. Therefore, how to make a tradeoff between the reporting channels
fading and the sensing diversity order need to be considered further.
Cooperative transmission technology with relay can effectively combat the
channel fading and enhance the system throughput [5, 57]. As stated in [10],
the relay technique has been generally considered as an effective method to improve
the capacity and coverage for next generation wireless networks, which is also
effective in cognitive radio networks. In [34], the authors have considered a secondary transmitter to act as a relay for primary transmissions. It has been shown that
the secondary link throughput can be improved in certain network topologies.
The CSS performance based on relay cooperation is investigated in [20] and [21],
which focus on how to cooperate between SUs. Paper [33] has investigated the use
of cooperative relay to assist the fulllment of heterogeneous trafc demands in a
secondary network with an unbalanced spectrum usage. They have proposed a
scheme to improve diversity by cooperative relay in CRN. More recently, papers
[5863] investigate the relay spectrum sensing protocal including multiple relay and
best relay, which mainly focus on the optimization of sensing overheads and the
enhancement of received signal by relay to improve the sensing performance. In the
research about the cooperative spectrum sensing with relay above, they mainly take
more attention on how to reduce the interference to the PU and other SUs. As to
how to improve the transmission reliability in the reporting channels, there is little
research, which needs to be in-depth study.
The system performance can be effectively improved by the soft combination
based CSS compared with the CSS based on hard combination. At present, the
research on the soft combination based CSS mainly consider the data fusion
method, in which, the SU can provide relatively detailed and effective local
detection information for the FC, and which has been studied in [48, 49]. The
authors proposed an optimal soft combination scheme and showed that the CSS
performance increased with the increase of the sensing users, however, the innite
bits are required and this will result in a large communication bandwidth when there
is large number of SU which will cause the great waste of communication bandwidth. A lot of work has been done on the quantization for the signal detection. In
[50], an optimal soft combination scheme is proposed, based on some approximation in the target optimal function. In [47], a linear soft combination of raw
measurements from individual cooperative SUs performed at the fusion center is
considered. All of the research consider the channel between SU and FC is perfect.
In fact, the imperfect reporting channels would induce the detection performance
degraded greatly [35, 64].
2.5 Research Status
39
The purpose of doing deep research to the spectrum sensing technology is to

make the CR apply to the actual wireless system more effectually. 3GPP and IEEE
either have put forward the some new requirements to the function of future
wireless networks, such as self-planning, self-establishment, self-deployment,
self-allocation, self-operation, self-optimization, self-healing and other features
[65, 66]. As to how to use the cognitive technology in wireless telecommunication
networks need to be further studied.
References
Proc IEEE 97(5):878893
2. Nosratinia A, Hunter TE, Hedayat A (2004) Cooperative communication in wireless networks.
3. Cover TM and El Gamal AA (1979) Capacity theorems for the relay channel. IEEE Trans Inf
Theor (IT-25): 572584
4. Kramer G, Gastpar M, Guptaz P (2005) Cooperative strategies and capacity theorems for relay
networks. IEEE Trans Inf Theory 51:30373063
5. Laneman JN, Tse DNC, Wornell GW (2004) Cooperative diversity in wireless networks:
efcient protocols and outage behavior. IEEE Trans Inf Theory 50(12):30623080
6. Sendonaris A, Erkip E, Aazhang B (2003) User cooperation diversitypart I: system
description. IEEE Trans Commun 51:19271938
cognitive radios. In: Proceedings of the 38th Asilomar conference signals, systems, computers.
8. Weiss T, Hillenbrand J, Jondral F (2003) A diversity approach for the detection of idle spectral
resources in spectrum pooling systems. In: Proceedings of the 48th international science
colloquium (Il-menau, Germany)
9. Hossain E, Bhargava V (2007) Cognitive wireless communication networks. Springer, New
York
10. Pabst R, Shultz BH (2004) Relay-based deployment concepts for wireless and mobile
broadband radio. IEEE Commun Mag 42(9):8089
11. Faccin SM, Wijting C, Kenckt J, Damle A (2006) Mesh WLAN networks: concept and system
design. IEEE Wireless Commun 13:1017
12. Oyman O, Laneman JN, Sandhu S (2007) Multihop relaying for broadband wireless mesh
networks: from theory to practice. IEEE Commun Mag 45:116122
environments. In: Proceedings of the 1st IEEE international symposium new frontiers dynamic
spectrum network DySPAN. Baltimore, MD, USA, pp 131136
cooperation-processing tradeoff. Wireless Commun Mobile Comput 7(9):10491060
16. Mishra SM, Sahai A, Brodersen RW (2006) Cooperative spectrum sensing among cognitive
radios. In: Proceedings of the 2006 IEEE international conference on communications, ICC
2006. Istanbul, Turkey, pp 16581663
17. Hillenbrand J, Weiss T, Jondral FK (2005) Calculation of detection and false alarm
probabilities in spectrum pooling systems. IEEE Commun Lett 9(4):349351
40
2 CSS Technology
18. Weiss TA, Hillenbrand J, Jondral FK (2003) A diversity approach for the detection of idle
spectral resources in spectrum pooling systems. In: Proceedings of the 48th international
science colloquium. Ilmenau, Germany
19. Ganesan G, Li YG (2005) B agility improvement through cooperation diversity in cognitive
radio. In: Proceedings of the IEEE global telecommunication conference (GLOBECOM), vol
5. St. Louis, MO, Nov 28Dec 2, pp 25052509
20. Ganesan G, Li Y (2007) Cooperative spectrum sensing in cognitive radio, part I: two user
21. Ganesan G, Li Y (2007) Cooperative spectrum sensing in cognitive radio, part II: multiuser
22. Visotsky E, Kuffner S, Peterson R (2005) On collaborative detection of tv transmissions
in support of dynamic spectrum sharing. In: Proceedings of the IEEE DySPAN 2005,
pp 338345
23. Gandetto M, Regazzoni C (2007) Spectrum sensing: a distributed approach for cognitive
terminals. IEEE J Sel Areas Commun 25(3):546557
24. Sendonaris A, Erkip E, Aazhang B (2003) User cooperation diversity part I: system
description. IEEE Trans Commun 51:19271938
25. Sendonaris A, Erkip E, Aazhang B (2003) User cooperation diversity Part II: implementation
aspects and performance analysis. IEEE Trans Commun 51:19391948
26. Nosratinia A, Hunter TE, Hedayat, A (2004) Cooperative communication in wireless
networks. IEEE Commun Mag, 7479
27. Ray Liu KJ, Sadek AK, Kwasinski A (2009) Cooperative communications and networking.
Cambridge University Press, New York
28. Hunter TE, Nosratinia A (2002) Cooperative diversity through coding. Proceedings of the
IEEE ISIT, Laussane, Switzerland 220
29. Hunter TE, Nosratinia A (2004) Diversity through coded cooperation, IEEE Trans. Wireless
Commun. 5(2):283289
30. Murch RD, Letaief KB (2002) Antenna systems for broadband wireless access. IEEE
Commun Mag 40(4):7683
31. Alamouti SM (1998) A simple transmit diversity technique for wireless communication.
IEEE J Select Areas Commun 16:14511458
32. Huang D, Letaief KB, Lu J (2004) A receive space diversity architecture for OFDM systems
using orthogonal designs. IEEE Trans Wireless Commun 3:9921002
33. Zhang Q, Jia JC, Zhang J (2009) Cooperative relay to improve diversity in cognitive radio
networks. IEEE Commun Mag 47(2):111117
34. Simeone O, Bar-Ness Y, Spagnolini U (2007) stable throughput of cognitive radios with and
without relaying capability. IEEE Trans Commun 55(12):23512360
in cognitive radio networks-[transaction letters]. IEEE Trans Wireless Commun 7(12):4761
4766
36. Digham FF, Alouini MS, Simon MK (2007) On the energy detection of unknown signals over
fading channels. IEEE Trans Wireless Commun 55(1):2124
37. Urkowitz H (1967) Energy detection of unknown deterministic signals. Proc IEEE 55(4):523
531
38. Kostylev VI (2002) Energy detection of a signal with random amplitude. In: Proceedings of
the 2002 international conference on communications (ICC 2002). New York, NY, United
states, pp 16061610
39. Digham FF, Alouini MS, Simon MK (2003) On the energy detection of unknown signals over
fading channels. In: Proceedings of the 2003 international conference on communications
(ICC 2003). Anchorage, AK, United States, pp 35753579
40. Simeone O, Gambini J, Bar-Ness Y et al (2007) Cooperation and cognitive radio. In: IEEE
international conference on proceedings of the communications. ICC07. Glasgow, pp 6511
6515
References
41
41. Scutari G, Palomar D, Barbarossa S (2008) Cognitive MIMO radio. IEEE Signal Process Mag
25(6):4659
42. Viswanathan R, Aalo V (1989) On counting rules in distributed detection. IEEE Trans Acoust
Speech Signal Process 37(5):772775
43. Quan Z, Cui S, Sayed AH (2007) An optimal strategy for cooperative spectrum sensing in
cognitive radio networks. In: Proceedings 50th annual IEEE global telecommunications
conference, GLOBECOM 2007. Washington, DC, United States, pp 29472951
44. Kam M, Zhu Q, Gray WS (1992) Optimal data fusion of correlated local decisions in multiple
sensor detection systems. IEEE Trans Aerosp Electron Syst 28(3):916920
45. Chen B, Jiang RX, Kasetkasem T et al (2004) Channel aware decision fusion in wireless
sensor networks. IEEE Trans Signal Process 52(12):34543458
46. Chair Z, Varshney PK (1986) Optimal data fusion in multiple sensor detection systems. IEEE
Transac Aerospace Electron Syst AES-22(1):98101
47. Zhi Q, Cui S, Sayed AH (2008) Optimal linear cooperation for spectrum sensing in cognitive
radio networks. IEEE J Selected Topics Signal Process 2(1):2840
48. Shen B, Kwak KS (2009) Soft combination schemes for cooperative spectrum sensing in
cognitive radio networks. ETRI 31(3):263273
49. Shen B, Kwak K, Bai ZQ (2009) Optimal linear soft fusion schemes for cooperative spectrum
sensing in cognitive radio networks. In: Proceedings of the 2009 IEEE global
telecommunications conference, GLOBECOM 2009. Honolulu, HI, United States, pp 16
50. Ma J, Zhao GD, Li Y (2008) Soft combination and detection for cooperative spectrum sensing
in cognitive radio networks. IEEE Trans Wireless Commun 7(11):45024507
52. Shen JY, Jiang T, Liu SY et al (2009) Maximum channel throughput via cooperative spectrum
sensing in cognitive radio networks. IEEE Trans Wireless Commun 8(10):51665175
53. Zheng SK, Liang YC, Kam PY et al (2009) Cross-layered design of spectrum sensing and
MAC for opportunistic spectrum access. In: Proceedings of the 2009 IEEE wireless
communications and networking conference, WCNC 2009. Budapest, Hungary, pp 16
54. Pei YY, Hoang AT, Liang YC (2007) Sensing-throughput tradeoff in cognitive radio
networks: how frequently should spectrum sensing be carried out? In: 18th annual IEEE
international symposium on personal, indoor and mobile radio communications, PIMRC07.
Greece, Athens, pp 18041805
55. Peh ECY, Liang YC, Guan YL et al (2009) Optimization of cooperative spectrum sensing in
cognitive radio networks: a sensing-throughput tradeoff view. IEEE Trans Veh Technol 58
(9):52945299
56. Zeng YH, Liang YC (2009) Spectrum-sensing algorithms for cognitive radio based on
statistical covariances. IEEE Trans Veh Technol 58(4):18041815
57. Hunter TE, Sanayei S, Nosratinia A (2006) Outage analysis of coded cooperation. IEEE Trans
Inf Theory 52(2):375391
58. Zou YL, Yao YD, Zheng BY (2010) Outage probability analysis of cognitive transmissions:
the impact of spectrum sensing overhead. IEEE Trans Wireless Commun 9(8):26762688
59. Zou YL, Yao YD, Zheng BY (2011) Cognitive transmissions with multiple relays in cognitive
radio networks. IEEE Trans Wireless Commun 10(2):648659
60. Zou YL, Yao YD, Zheng BY (2011) A cooperative spectrum sensing based cognitive relay
transmission scheme without a dedicated sensing relay channel in cognitive radio networks.
IEEE Trans Signal Process 59(2):854858
61. Zou YL, Yao YD, Zheng BY (2011) A selective-relay based cooperative spectrum sensing
scheme without dedicated reporting channels in cognitive radio networks. IEEE Trans
Wireless Commun 10(4):11881198
62. Zhu J, Zou YL, Zheng BY (2009) Cooperative detection for primary user in cognitive radio
networks. EURASIP J. Wireless Commun Netw. doi:10.1155/2009/617320 [Online].
Available: http://downloads.hindawi.com/journals/wcn/2009/617320.pdf
42
2 CSS Technology
63. Zou YL, Zhu J, Zheng BY et al (2010) An adaptive cooperation diversity scheme with
best-relay selection in cognitive radio networks. IEEE Trans Signal Process 58(10):54385445
cognitive radio networks. In: IEEE international conference on communications. China,
pp 34113415
65. Shen J (2004) 3GPP long term evolution: principle and system design, Beijing. The Peoples
Posts and Telecommunications Press, Beijing
66. Stefania S, Issam T, Matthew B (2010) LTEthe UMTS long term evolution from theory to
practice, Beijing. The Peoples Posts and Telecommunications Press. Beijing
Chapter 3
The Relationship Among the Performance

Parameters in CSS
In this chapter, the basic system model of CSS is illustrated. The fundamental
detection performance and secondary throughput are given. Most of all, the relationship of the parameters in CSS are analyzed in detail, which can help readers
understand preferably.
3.1
Introduction
In CRN, the SU can utilize the spectrum that the PU has not occupied to improve the
spectrum utilization [15]. There are two parameters associated with spectrum
sensing: probability of detection and probability of false alarm. The higher the
probability of detection, the better the PU is protected. However, from the SUs
perspective, the lower the probability of false alarm, the more chances the channel
can be reused when it is available. In actual occasion, the main metric of sensing
performance is: (1) minimizing the false alarm probability for a target detection
probability; (2) maximizing the detection probability under given target false alarm
probability [6]. The system performance gain caused by users collaboration has
been investigated when sensing channel is AWGN channel in [7, 8], in which, the
two users cooperation method and multiple-users cooperation method are given
based on the periodic spectrum sensing. Researchers have studied the affect of
sensing diversity order to the CSS performance when the sensing channel is AWGN
channel [911]. The result shows that the performance of CSS is improved with the
increase of sensing diversity. Furthermore, it has been proved theoretically that
compared with the single user spectrum sensing, the CSS reduce the demand of the
average signal-to-noise spectrum sensing channel. In [1216], the authors have
studied the relationship between the sensing time and the secondary throughput.
A longer sensing duration could produce a better sensing performance but results
in longer waiting time for SUs to access the channel, thus the lower spectrum
utilization. For this, An optimal sensing duration has been studied to improve the
DOI 10.1007/978-3-319-31095-4_3
43
44
3 The Relationship Among the Performance Parameters in CSS
system performance from the point of secondary throughput in [1720], in which, the
TDMA (Time Division Multiple Access) protocol was considered for sending local
sensing results. Obviously the fewer SUs involved in CSS, the shorter the whole
sensing duration. However, a small number of SUs in CSS results in a small sensing
diversity order which will lead to the lower probability of detection. This problem
can be addressed by allowing the SUs to send the decisions on orthogonal frequency
bands [2123]. In [2123], the authors have pointed out that, when sending the local
decisions on orthogonal frequency bands, a large portion of the available bandwidth
will be required with the increase of sensing users. Therefore, how to reduce the
system resources consumption while guaranteeing the good sensing performance as
soon as possible needs to be further studied. Accordingly, in [24], the authors have
studied the best number of sensing users for minimizing the system total error
probability in spectrum sensing, and realize a tradeoff between the system detection
performance and the control bandwidth. The system total error probabilities consist
of two parts: the false alarm probability and the miss detection probability. The lower
the miss detection probability is, the higher the detection probability. In the actual
application, it will be not realizable when satisfying one of the parameters (i.e. the
false alarm probability or the miss detection probability) to optimize the other no
matter what decision rules are utilized according to the method as in [24].
The CSS includes two successive processes: the sensing process and the
reporting process. In the actual wireless environment, the hidden terminal problems
exist not only in sensing channels but also in reporting channels. The reporting
channels fading will make influence to CSS performance [2123].
In this chapter, we consider that the local sensing results are transmitted to the
FC in OFDMA mode. Under the actual occasions of both sensing channels and
reporting channels are fading channels, we analyzed the relationship among the
number of SU, the detection threshold, the bandwidth, the spectrum utilization and
the secondary throughput, as described follows:
Firstly, the influences of sensing diversity order to the secondary throughput.
In CSS, higher sensing diversity order makes more chances for SUs to detect the
vacant spectrum and access it, which may make the secondary throughput
improved. While on the other hand, the higher sensing diversity order the more
control bandwidth occupied for reporting the local sensing results and the lower
available bandwidth for SUs transmission, which will make the secondary
throughput reduced. So how to select the appropriate SU to maximize the secondary
throughput deserved further research.
Secondly, the relationships among the detection threshold, detection performance and the secondary throughput. In CSS, the detection performance is affected
by the detection threshold directly. The lower the detection threshold, the better the
detection performance, which results the PUs being protected better. However,
lower detection threshold also results in higher probability of false alarm, which
make less chances for SUs to access the channel and so the reduced secondary
throughput. So how to design the appropriate detection threshold to achieve the
maximized secondary throughput while ensuring sufcient protection to the PU
should be studied further.
3.2 System Model
3.2
45
System Model
The CSS process based on hard combination is shown in Fig. 3.1, in which, both
the sensing channels and reporting channels are undergoing fading. The original
status of the PU is assumed as H0 or H1, representing the vacant spectrum present or
not. In general, the process of CSS can be divided into three steps as follows:
1. Every SU i performs local spectrum measurements independently and then
makes a binary decision Di 2 {0, 1} for all i = 1, , N. Specically, {0}
indicates that the SU infers the absence of the PU in the observed band. In
contrast, {1} infers the operating of the PU.
2. All of the SUs forward their binary decisions to a fusion center (FC) via the
reporting channels. Here, we consider the reporting channels as fading channels.
3. At the FC, all 1 bit decisions are fused together according to certain fusion rules
and a nal decision are made. In this paper, we adopt the k-out-n rule.
In the CSS, when the local sensing results are transmitted to the FC by TDMA
mode, only one of the SUs are allowed to forward its local decision results to the
FC in each time slot. When the number of the local decision value N increases, the
total reporting time for sending the local decision results sR will increase as shown
in Fig. 3.2, which results the increase of the total sensing duration. In the periodic
spectrum sensing, the SU may not vacant the channel immediately when the PU
returns if the sensing duration is long, which will make interference to the primary
signal [21]. To assure sufcient protection to the PU, we utilize OFDMA mode for
transmission, i.e. each SU sends its initial detection result to the FC on orthogonal
frequency bands to solve the interference problem induced by longer sensing
duration [7, 8, 21], as shown in Fig. 3.3. The bandwidth resources are measured by
subbands.
s (t )
PU
H0 /H1
1th sensing
channel
ith sensing
channel
Nth sensing
channel
x1 (t )
SU1
xi ( t )
SUi
xN ( t )
SUN
D1
1th reporting
channel
Di
ith reporting
channel
DN
Nth
reporting
channel
Fig. 3.1 The CSS process under both imperfect channels
Time slot 1
Time slot 2
Time slot N
R
Fig. 3.2 The sending mode of local sensing results-TDMA
D
1
D i
D N
FC
46
subband1 subband2
subbandN
B
Fig. 3.3 The sending mode of local sensing results-OFDMA
3.3
3.3.1
The CSS Detection Performance

Local Detection Performance
The spectrum sensing can be regarded as a binary hypothesis testing problem:

(
H0 : PU is absent
H1 : PU is in operation
The key parameters of the spectrum sensing include the probabilities of error
detection and the probabilities of correct detection. Considering the process of the
CSS in Fig. 3.1, we suppose that there are N independent cognitive radio users
named as SU in cognitive radio system, whose coverage is far less than the distance
between cognitive radio network and the primary network.
When ith SU performs local spectrum sensing independently, xi t is the
observed signal at the ith SU at time t. The local spectrum sensing problem can be
described as the following hypothesis:
(
xi t
ni t; H0
hi st ni t; H1
3:1
where, ni t is the additive white Gauss noise (AWGN) in the sensing channel of ith
SU, st is the signal from the PU, hi is the channel gain of the sensing channel
between the PU and the ith SU.
When the ith SU make a determination based on the local observation, the
energy collected in the frequency domain is denoted by Yi . The observation
bandwidth is W and the observation time window is T. Then Yi obeys the following
distribution [21]:
(
Yi
v22u ; H0
v22u 2ci ; H1
3:2
where, u = TW is the time bandwidth product, v22u denotes a central chi-square

distribution with degrees of freedom 2u, v22u 2ci represents a noncentral chi-square
distribution with degrees of freedom 2u, ci is the instantaneous SNR of the received
3.3 The CSS Detection Performance
47
signal at the ith SU. The detection of PU signals can be realized by comparing the
energy Yi with a predened threshold ki , the rules are as follows:
(
Yi
[ki ; H1 :
denotes the observed spectrum busy, let Di 1
\ki ; H0 : denotes the observed spectrum idle, let Di 0
3:3
Here, we assume that, compared with the distance from any SU to the primary
transmitter, the distance between any two SUs is small, so that the received signal at
each SU experiences almost identical pathloss. Therefore, it is reasonable to assume
that the primary signals received at the SUs are independent and identically distributed (i.i.d.) with the instantaneous SNRs in every link between PU and SUs the
same mean c. Furthermore, we assume that all SUs use the same threshold and the
noise power is independent and identically distributed. Therefore, the probability
density function of Y under the two hypothesis can be written as follows [25, 26]:
8
>
>
>
<
y
yu1 exp ; H0
2
fY y
u1=2

p
>
1 y
2c y
>
>
2cy ; H1
exp
Iu1
:
2 2c
2
1
2u Cu
3:4
Over AWGN channels, the local false alarm probability ^pf and the detection
probability ^
pd can be calculated as [21]:
Z1
^
pf PfY [ kjH0 g
fY jH0 ydy
k
Cu; k=2
C u
Z1
^
pd PfY [ kjH1 g
fY jH1 ydy Qu
p p
2c; k
3:5
3:6
where, Ca is the Rgamma function, Ca; x is the incomplete gamma function

1
given by Ca; x x ta1 et dt, Qm a; b is the generalized Marcum Q-function
R1 m
x2 a2
expressed as Qm a; b b axm1 e 2 Im1 axdx, Iv is the vth order modied
Bessel function of the rst kind.
It can be obtained that the SUs local miss detection probability can be calculated
as ^
pm 1 ^
pd . The higher miss detection probability means that there is higher
probability to infer the absence of the PU, which may cause interference to the PU.
However, the higher false alarm probability means that the presence of the PU is
miss detected which will result the lower access opportunities for SU and the lower
spectrum utilization.
48
From Eq. (3.5), ^pf is independent of c because under H0 there is no PU signal

present. When hi is varying due to shadowing or fading, the average false alarm
probability and the detection probability of SU can be calculated as follows [21]:
8
P ^pf
>
>
< f
Z
^pd cf cdc
P
d
>
>
:
3:7
where, f c is the pdf of SNR in sensing channels. The Rayleigh fading sensing
channels are considered in this paper, whos SNR obeys the exponential
distribution:

1
c
f c exp ; c 0
c
c
3:8
where, c is the average SNR of the sensing channels.

Under Rayleigh fadings, the detection probability of SU can be calculated as
[10]:
Z
Pd
c
3.3.2
u2
X
1
k=2n 1 c=cu1
n!
n0
"
#
u2
X
1
n
k=21 c
k=2
kc=21 c
e
e
n!
n0
^pd cf cdc ek=2
3:9
CSS Performance Based on Decision Fusion
According to the process of CSS described above, each SU make a 1bit local
decision Di based on energy detection, and then forward it to the FC via a reporting
channel. Here, we consider the reporting channels as fading channels, which will
cause the informations error transmission. We suppose that the probability of
reporting error is identical for all SUs as Pe. The binary hypothesis decisions made
by SU denoted as H1 and H0 , which represent the observed band is occupied or
idle. All of the local decisionsconstitute a set
represented as D fD1 ; D2 ; . . . DN g,
and the set denoted as D0 D01 ; D02 ; . . . D0N is the values after the transmission
through the reporting channels. Then, before the local decisions are transmitted to
the FC through reporting channels, the detection probability and the false alarm
probability of each SU can be calculated as following [27]:
49

P f Pr H1 H0

Pr H1 H0 Pr H1 H1 Pr H0 H0 Pr H1 H0

Pf 1 Pe 1 Pf Pe
3:10
P d PrH1 jH1
PrH1 jH1 PrH1 jH1 PrH0 jH1 PrH1 jH0
Pd 1 Pe 1 Pd Pe
3:11
where, PrBj A denotes the probability for making estimation to the event A as the
event B.
Finally, the FC combines all of the decisions and makes nal decision as H1 and
H0 denoted the observed band being occupied or idle. The fusion criterion of the
FC is [21]:
F
N
X
(
^i
D
K; H1
\K; H0
i1
3:12
According to the fusion rule in Eq. (3.12), it is seen that the total false
alarm probability Qf and the detection probability Qd of CSS can be calculated as
[11, 28, 29]:
Qf PrfH1 jH0 g
Qd PrfH1 jH1 g
N
X
iK
N
X
iK
!
PrH1 jH0 i PrH0 jH0 Ni
3:13
!
PrH1 jH1 i PrH0 jH1 Ni
3:14
According to Eqs. (3.10), (3.11), (3.13) and (3.14) can be further calculated as:
Qf
Qd
N
X
iK
N
X
iK
Ni
Pif 1 Pf
3:15
!
Pid 1 Pd Ni
3:16
where, K 2 f1; 2; . . . ; N g; when K = 1, the k out of n rule becomes or rule;

when K = N, the k out of n rule becomes and rule. The interference to the PU can
be reduced utilizing or fusion rule which is a protective strategy for the SU. The
following simulation results also show that the or fusion rule is relatively an
optimized criterion.
50
3.3.3
Simulation Results
3.3.3.1
Single Spectrum Sensing Performance
Figure 3.4 shows the single spectrum sensing performance when the sensing
channel is AWGN channel with the average SNR is 0, 5 and 10 dB respectively.
The time-bandwidth product is u = 5. The abscissa is the false alarm probability, the
ordinate is the miss detection probability. It is seen that the miss detection probability varies with the SNR when the false alarm probability is given. The larger the
SNR, the smaller the miss detection probability, which means that the PU signal is
detected more accurately.
Figure 3.5 illustrates the single spectrum sensing performance when the sensing
channel is Rayleigh fading channel with the average SNR is 0, 5 and 10 dB
respectively. The abscissa is the false alarm probability; the ordinate is the miss
detection probability. Comparing Figs. 3.4 and 3.5, we can see that the signal point
spectrum sensing under Rayleigh fading channel is worse than that of AWGN
channel. For example, when the false alarm probability is 0.1, the miss detection
probability is 0.99, 0.9 and 0.12 under Rayleigh fading channel with the average
SNR is 0, 5 and 10 dB respectively, while which is 0.8, 0.12 and 0.04 under
AWGN channel. This is mainly due to that the signal received by the SU is too
weak caused by the channel fading, which results in the bad detection performance.
10
the miss detection probability
-1
10
-2
10
-3
10
-4
10
-5
10
SNR=0dB
SNR=10dB
SNR=20dB
-6
10
-2
10
-1
10
the false alarm probability
Fig. 3.4 The detection performance under AWGN channels
10
51
10
-1
10
-2
10
-3
10
SNR=0dB
SNR=5dB
SNR=10dB
-4
10
-2
10
-1
10
10
Fig. 3.5 The detection performance under Rayleigh channels
3.3.3.2
The CSS Performance Under Imperfect Sensing Channels
According to the results of the single spectrum sensing performance in

Sect. 3.3.3.1, the spectrum sensing performance is too poor when the sensing
channels are Rayleigh fading channels. The previous theoretical analysis has shown
that, the detection performance under Rayleigh fading channel can be effectively
improved by multi-user collaborative spectrum sensing. In this section, we will
illustrate the CSS performance curve when the sensing channels are Rayleigh
fading channel, in which, the average SNR is 10 dB in the sensing channels, the
time-bandwidth product u = 5.
Figure 3.6 illustrates the CSS performance using different fusion criterion, in
which, the SU number in cooperation N = 10. It is seen that the lower miss
detection probability can be achieved by the or fusion rule, which means the better
detection performance. The research in literature [6] has shown that, in many cases,
the better detection performance can be achieved by the or fusion rule compared
with other fusion rules. Therefore, we use or fusion rule throughout this paper in
the decision fusion algorithm.
Figure 3.7 shows the CSS performance under different sensing users N use the
or fusion rule. It is seen that, for a given false alarm probability, the miss detection
probability obviously decreases with the increasing of the sensing users N. This is
mainly due to the multi-user diversity with multiple SUs which makes the better
detection performance.
52
10
-1
10
-2
10
-3
10
k=1('or')
k=5
k=10('AND')
-4
10
-3
-4
10
10
-2
10
-1
10
10
Fig. 3.6 The CSS performance using k out of n rule when N = 10
10
-1
10
-2
10
-3
10
N=1
N=2
N=5
N=10
-4
10
-4
10
-3
10
-2
10
Fig. 3.7 The CSS performance under different sensing users
-1
10
10
53
10
-1
10
-2
10
-3
10
Pe=0.1
Pe=0.01
Pe=0.001
Pe=0
-4
10
-4
10
-3
-2
10
10
-1
10
10
Fig. 3.8 The CSS performance under different reporting error probabilities
3.3.3.3
The CSS Performance Under the Imperfect Reporting

Channels
Figure 3.8 illustrates the CSS performance with the Rayleigh fading sensing
channels and the reporting error probability as 101, 102, 103 and without error in
reporting channels, in which, the simulation parameters are given by SNR = 10 dB,
u = 5, N = 10. It can be seen that there exists a lower bound of the false alarm
probability under the imperfect reporting channels, i.e. the error probabilities in
reporting channels are not equal to zero. At the lower bound of the false alarm
probability, the miss detection probability will drastically increase. Moreover, this
lower bound increases with the reporting error probabilities increase. However, the
larger false alarm probability means the less chance for SU to access and the lower
detection performance, which is mainly due to the local sensing informations error
transmission by the reporting channels fading.
3.4
3.4.1
The CSS Secondary Throughput

Spectrum Utilization
In CRN, the SUs are allowed to share the spectrum originally allocated to the PU
opportunistically to realize dynamic spectrum access as long as the vacant spectrum
54
are detected. While in practice, the sensing performance is often restricted by

multipath fading, shadowing effect and the hardware conditions of the receiver,
thereafter, the SUs may not be able to accurately detect whether the PU signal exists
or not. In some cases, even though there is idle spectrum in the network, it may not
be utilized by the SUs when the SUs have yet mistakenly detected the PU signal.
The percentage of the correctly identied vacant spectrum provides a reasonable
measure of the spectrum utilization, which is determined by 1 Qf . The smaller Qf
leads to the higher percentage of correctly identied vacant spectrum, which
resulting the improved spectrum utilization. From Sect. 3.3, we can see that Qf is
related to the local detection threshold and the sensing diversity order. In what
follows, we will show how the sensing diversity order and local detection threshold
make closely influence to the false alarm probability of CSS.
According to Eqs. (3.5), (3.7), (3.10) and (3.13), the following formulas can be
obtained:
u1
@Pf
1
k
k

e2 \0
2Cu 2
@k
3:17
@P f
1 2Pe 0
@Pf
3:18
NK
N1
[0
P Nf 1 P f
K1
3:19
@Qf
N
@P f
According to the analysis in Sect. 3.3, in practice, if the reporting channel

condition is too bad, the CSS performance is so poor that it is no meaningful
actually. In CSS, we should avoid the occasions that the reporting channels are too
poor through various methods, which will be studied in-depth in the third chapter
and the fourth chapter latter. Here, we assume that the reporting error is less than
0.5 in actual occasions. From Eqs. (3.17)(3.19), we get
@Qf @Qf @P f @Pf
\0

@k
@P f @Pf @k
3:20
The local detection probability under AWGN channels in Eq. (3.6) can be
expressed as:
Pd
AWGN
2

Z1
p
1
x 2c
u
2cx dx
p u1
x exp
Iu1
2
2c p
k
3:21
3.4 The CSS Secondary Throughput
55
Then, we can get

@Pd
AWGN
@k

u1
p
k2
k 2c
exp

2ck \0
I
u1
p u1
2
2 2c 2
3:22
The expression of the detection probability under Rayleigh fading sensing

channels is relatively complex, which may be derived by averaging the detection
probability under AWGN sensing channels over fading channel statistics.
Therefore, the variation trends of the detection probability vs. the local detection
threshold under the two channels are the same. Combining Eqs. (3.7) and (3.8), we
can get
@Pd
Rayleigh
@k
@Pd
AWGN
@k
fc xdx\0
3:23
From Eqs. (3.11) and (3.16), the following formulas can be obtained:
@P d
@Pd
@Qd
N
@P d
1 2Pe [ 0
3:24
!
N1
P Nd 1 P d NK [ 0
K1
3:25
Rayleigh
According to Eqs. (3.23)(3.25), the variation of the CSS detection probability

with the local detection threshold k can be expressed as:
@Qd @Qd
@P d
@Pd Rayleigh
\0

@k
@P d @Pd Rayleigh
@k
3:26
According to the above analysis, the detection probability and the false alarm
probability of CSS are both monotonically decreased with the detection threshold.
When N is given, the higher the detection threshold is, the lower the false alarm
probability and the higher the spectrum utilization, whereas the detection probability decreases.
According to the characteristics of the summarized function, the verdict that Qf
and Qd monotonically increase with the sensing diversity order could be easily
obtained, that is
@Qd
[0
@N
3:27
@Qf
[0
@N
3:28
56
From the above analysis, it is found that Qf is influenced by sensing diversity

order N and detection threshold . When was given, the smaller N leads to the
lower Qf and the higher spectrum utilization. On the other hand, when N was given,
the higher results in the lower Qf and the higher spectrum utilization, whereas, the
detection probability Qd decreased which will bring interference to the PU. So,
there exists tradeoff between spectrum utilization and protection level to the PU,
which is affected by sensing diversity order and detection threshold.
3.4.2
Secondary Throughput
In CSS, the larger sensing diversity order makes the better detection performance
and more chances of vacant spectrum to be found, which results in the higher
spectrum utilization and the improved secondary throughput. While on the other
hand, the larger sensing diversity order leads to more control bandwidth occupied
for reporting local sensing results to the FC, which would induce the decreased
secondary throughput because of the reduced bandwidth for transmitting data as
shown in Fig. 3.9, where, W is the total system bandwidth, B represents the control
channel bandwidth occupied by sending 1bit local decision. Obviously, in actual
occasions, it is required that W NB 0. What is needed to be explained is that we
mainly concern the effect of the control bandwidth to the system performance in
sending the local sensing results in this paper. The bandwidth required by the SUs
for performing the local detection will not affect the performance analysis in this
paper, so we do not care this part of bandwidth in the following.
For a given frequency band of interest, let us dene P1 as the probability of the
PU being active, and P0 as the probability of the PU inactive. Only when the PU is
inactive the SU can access the licensed spectrum to communicate. While in practice, the PU may be present or not, and the missing detection happens with the
probability of 1 Qd when the PU is present, the correct detection happens with
the probability of 1 Qf when the PU is absent. Therefore, there are two scenarios
for which the secondary network can operate at the PUs frequency band.
Scenario I: When the PU is not present and no false
alarm is generated by the

SU, the probability for which happens is P0 1 Qf . The achievable
throughput of the SU is WNB
d0 . Where, d0 is the throughput of the secondary
W
network when the channels are idle.
Scenario II: When the PU is active but it is not detected by the SU and the
probability for which happen is P1 1 Qd . The achievable throughput of the
Fig. 3.9 The bandwidth
allocation in CSS
57
SU is WNB
d1 . Where, d1 is the throughput when it operates in the presence
W
of PU.
According to the analysis in Sect. 2.3, Qf and Qd are both the function of N and
k, the average throughput of the SU in the above two kinds of scenarios can be
dened as follows:
C0 N; k

W NB
d0 P0 1 Qf N; k
W
3:29
C1 N; k
W NB
d1 P1 1 Qd N; k
W
3:30
Then, the average throughput CN; k for the secondary network can be
expressed as:
CN; k C0 N; k C1 N; k
3:31
It is obvious that the smaller sensing diversity order, the less control bandwidth
occupied for reporting the local sensing results and the larger available bandwidth
for SUs transmission, which will make the secondary throughput increased. While
on the other hand, according to the preceding analysis, we can know that the
detection probability and the false alarm probability of CSS are both monotonically
decreased with the detection threshold, increased with the sensing diversity order.
In actual occasion of spectrum sharing, what we care is how to increase the
spectrum utilization opportunity for SU under PU is protected sufciently, so as to
improve the secondary throughput.
d , the rst
From Eqs. (3.29)(3.31), for a given target probability of detection, Q
item in the right hand side of Eq. (3.31) dominates the achievable throughput.
Therefore, according to reference [17], for convenience, the normalized achievable
throughput could be written as:

~ N; k 1 Na 1 Qf N; k
C
3:32
where, = B/W is the ratio of the occupied control bandwidth for sending 1 bit local
decision to the overall system bandwidth.
Combing Eqs. (3.20) and (3.28) with Eq. (3.32), the following can be obtained:
~ N; k
@C
@Qf N; k
1 Na
[0
@k
@k
3:33
~ N; k

@C
@Qf N; k
a 1 Qf 1 Na
\0
@N
@N
3:34
58
The above analysis indicate that for a given target probability of detection, the
smaller the sensing diversity order is, the less the control bandwidth will be occupied
and the more available bandwidth for transmitting, which corresponding to the
higher secondary throughput. While on the other hand, there is a necessity of lower
detection threshold to satisfy the target detection probability once the sensing
diversity order reduced, which will lead to the increased probability of false alarm in
reverse. Therefore, how the secondary throughput changed is hard to be testied in
theory since the influence of the sensing diversity order and the local false alarm
probability to the CSS false alarm probability is inverse. We should nd a tradeoff
between the sensing diversity order and the detection threshold to maximize the
secondary throughput while assuring sufcient protection to the PU.
3.4.3
The Optimal Algorithm
3.4.3.1
N-P Criterion
In the cognitive systems, the two detection performance parameters are the detection probability of PH1 jH1 and the false alarm probability of PH1 jH0 . We hope
that the correct detection probability of PH1 jH1 is as large as possible, and the
false alarm probability of PH1 jH0 is as small as possible. However, when
the signal-to-noise ratio is given, the value of PH1 jH1 and PH1 jH0 grow with
the same trend, i.e. when PH1 jH1 increases PH1 jH0 will also increase.
Therefore, the Neyman-Pearson criterion was proposed denoted as N-P criterion
[30]: maximizing the correct detection probability of PH1 jH1 under the constraint
of false judgment probability PH1 jH0 a. Much false data are avoided and the
useful data are saved by the N-P criterion. In reference [31], the authors described
the existence of solutions to the N-P criterion from concept, they show that there
exists at least a kind of decision region segmentation, which can not only ensure the
condition of PH1 jH0 a, but also can make the values of PH1 jH1 maximized.
3.4.3.2
Algorithm Design
According to the above analysis, the optimization problem of the secondary

throughput can be described as: nding the optimal pair of N; k to maximize the
secondary throughput while maintaining sufcient protection to the PU. From the
perspective of the detection performance parameters, the protection level to the PU
is mainly reflected through the detection probability, the higher the detection
probability, the stronger protection can be achieved. When the detection probability
is equal to 100 %, the PU is absolutely non-disturbed, while the SUs can not access
59
to the licensed spectrum. In practice, in order to realize the reuse of the spectrum,
we only make sure that the PU got enough protection. Consequently, the optimization problem of the secondary throughput can be described as: nding the
optimal pair of N; k to maximize the secondary throughput while satisfying the
d , where, Q
d is the minimum target detection probability,
condition of Qd Q
representing the minimum protection level to the PU, whose value is generally
between 0.9 and 1. From Eq. (3.31), the secondary throughput is the function of the
sensing diversity order and the detection threshold. According to the N-P criterion
[29, 30], the optimization problem can be expressed as [32, 33]:
8
< max C N; k C0 N; k C1 N; k
N;k
: s:t: Q Q
d
d
3:35
d is given, the item of C0 N; k

When the target detection probability of Q
dominates C N; k, then, for a given diversity order, the optimization problem is
transformed into the following constraints:
(

~ k 1 Na 1 Qf k
max C
k
s:t:
d
Qd Q
3:36
For a given sensing diversity order of N, we choose a detection threshold k0 such

d . We may also choose an arbitrary detection threshold making
that Qd k0 Q
k1 k0 such that Qd k1 Qd k0 . Obviously, Qf k1 Qf k0 . Thus we have
~ k1 C
~ k0 . Therefore, the optimal solution to Eq. (3.36) is achieved with
C
equality constraint in Eq. (3.36). In another way, the maximized secondary
throughput can be achieved in maintaining PU sufciently protected when k0 is
selected as detection threshold.
In the traditional algorithm, the optimal detection threshold was obtained by
simulation. First, they nd all of the target value(the detection probability or the
false alarm probability) corresponding to all the detection thresholds, and then
comparing them one by one to get the optimal detection threshold which makes the
value maximize or minimize. Although this method can be implemented in the
background, it is not convenient in practical applications. The optimization algorithm proposed in this paper can be directly obtained by solving the mathematical
model to get the optimal detection threshold, which has more flexibility and
applicability.
Steps of the optimal algorithm proposed are elaborated as follows:
1. For each sensing diversity order of N, it is need to nd the optimal detection
threshold by some methods. Based on the analysis above, several existing
algorithms (Newton-Raphson, Bi-Section, SECOND, etc.) may be used to nd
60
d . However, the Newton-Raphson method

the root of the equation Qd k Q
needs the explicit expression of derivative of function, and it has faster converging rate than the other algorithms [18, 28, 30]. Since the explicit expression
of the detection probability to the detection
threshold
exists, the Newton-Raphson

d . The process of the
method is preferable here. Let g kN Qd kN Q
Newton-Raphson algorithm is listed as follows
Step 1 choose
and the initial guess kN 1, let i = 1;

tolerance

Step 2 if g kN i \e, stop; otherwise, go to step 3

Step 3 let kN i 1 kN i g kN i =g0 kN i . Then, let i = i+1 and go
to step 2.
Through the above three steps, we can get the optimal detection threshold for
maximizing the secondary throughput when the target detection probability is
satised. It is needed to note that we should combine Eqs. (3.6), (3.11) and
(3.14) to get the above solution.
2. According to the Eqs. (3.5)(3.7) and (3.9), calculating the corresponding local
opt
opt
probability of false alarm Popt
kN and the probability of detection Popt
kN
f
d
under Rayleigh fading sensing channels.
opt
kN according
3. Calculate the corresponding false alarm probability of CSS Qopt
f
to the Eq. (3.13).

~ opt kopt
4. Calculate the optimal normal secondary throughput C
according to the
N
Eq. (3.32).
3.4.4
Simulation Results
To get insight into the effectiveness of the proposed sensing methods and validate
some related theorems, extensive computer simulations have been conducted in this
section.
Figure 3.10 illustrates some results of CSS performance with different number of
sensing users N, in which sensing parameters are given as K = 1, Pe = 0.1, P0 = 0.5,
SNR = 10 dB, = 0.1. The detection thresholds are normalized. It is seen that, for a
given detection threshold , as the number of sensing users N increases, the
detection probability and false alarm probability of CSS both increase which
attribute to the cooperation of multiuser, however, the secondary throughput
decreases. This is because the enhanced performance of secondary throughput
brought by the increase of access opportunity could not compensate for the performance loss induced by the consumed bandwidth.
61
(a)
N=1
N=2
N=4
N=8
N=10
Probability of false alarm
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Normal throughput of second system
(b)
10
15
20
25
20
25
30
Threshold
0.9
N=1
N=2
N=4
N=8
N=10
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
10
15
30
Threshold
(c)
Probability of detection:Qd
0.9
0.8
0.7
0.6
0.5
0.4
0.3
N=1
N=2
N=4
N=8
N=10
0.2
0.1
10
20
30
40
50
60
Threshold
~
Fig. 3.10 The performance of CSS with different SU number. a Qf versus detection threshold. b C
versus detection threshold. c Qd versus detection threshold
62
While for a given N, when the detection threshold increases, the false alarm
probability decreases, whereas the secondary throughput increases. i.e. the larger
the detection threshold is the higher the secondary throughput. However, from
Fig. 3.10c we can see that the detection probability decreases with the increasing of
the detection threshold. It is obvious that there exists corresponding best detection
threshold for different number of sensing users when the target detection probability
d 0:9 is satised, and a larger value of N requires higher detection threshold. If
Q
a detection threshold could make the secondary throughput maximized under certain value of N when satisfying the target detection probability, the same object of
other values could not be satised nevertheless. For example, k 1 is the optimal
detection threshold when N is 1, in which the maximized secondary throughput can
be achieved while assuring the detection probability of CSS being 0.9. When k [ 1,
although the secondary throughput increases, the detection probability is less than
0.9. It is obvious that k 1 is not the optimal detection threshold when N = 2. As a
demonstration, the optimal detection threshold and corresponding detection probability, secondary throughput, as well as other detection probability and secondary
throughput are given in Table 3.1 when N is 1, 2 and 4 respectively. Therefore, for
the different number of cooperative users N, there exists different optimal detection
threshold to maximize the secondary throughput while ensuring sufcient protection to the PU. What needed to explain is that, when the local detection threshold is
greater than 30, the false alarm probability curve and the secondary throughput
curve show a gradual trend, while the detection probability is still decreased. Thus,
in Fig. 3.10, we only give the false alarm probability and the secondary throughput
curve as the detection threshold value is within 30, and the detection probability is
required to give the curve of detection threshold within 60 in order to analyze the
performance of CSS for satisfying the target detection probability of 0.9.
In the Sect. 3.4.3, the optimal algorithm has been given. We will illustrate the
effectiveness of the optimal algorithm from the point of the simulation results as
shown in Fig. 3.11 as an example in the following. Figure 3.11 shows the CSS
performance using optimal algorithm under different average signal and noise ratio
d 0:9.
of sensing channels (SNR), in which, K = 1, Pe = 0.1, P0 = 0.5, = 0.1, Q
Table 3.1 The optimal detection threshold and corresponding performance with different SU
number
The normalized
secondary throughput
and detection probability
~
=1
C
The number of SU
N=1
N=2
N=4
0.1
0.02
= 15
Qd
~
C
0.9
0.72
0.98
0.5
1
0.25
= 27
Qd
~
C
0.68
0.8
0.9
0.62
0.98
0.38
Qd
0.46
0.68
0.9
63
Normal throughput of secondary system
(a) 0.7
SNR=0dB
SNR=5dB
SNR=10dB
0.6
0.5
0.4
0.3
0.2
0.1
0
1
10
The number of secondary user: N
(b)
SNR=0dB
SNR=5dB
SNR=10dB
0.9
0.8
0.7
0.6
0.5
0.4
10
The number of secondary user:N
Fig. 3.11 The performance of CSS under various sensing channels using the proposed method.
a The normal secondary throughput versus the number of sensing users. b Qf versus the number of
sensing users
The normal secondary throughput versus the number of sensing users is illustrated
in Fig. 3.11a, in which, when SNR = 10 dB and N = 2 the normalized secondary
d 0:9 the
throughput is 0.5. On the other hand, from Fig. 3.10c, when N = 2 and Q
corresponding detection threshold is 15, and from Fig. 3.10b we can see that when
= 15 and N = 2, the normalized secondary throughput is also equal to 0.5, which
validate the optimal algorithm.
64
From Fig. 3.11a, it is seen that, the secondary throughput increases with SNR
increasing for a given N. This is reasonable. In fact, in order to achieve the target
detection probability of CSS, the larger the SNR is, the higher detection threshold is
required, and the smaller the target false alarm probability of CSS is, which make
the secondary throughput increases. It is also seen that, the secondary throughput
increases rst, decreases then with the increase of the number of sensing users. This
is due to the effect of diversity reception. Because of the diversity, the performance
of CSS outperforms the single spectrum sensing when N increases from 1 to 2, i.e.
the false alarm probability Qopt
decreases greatly while the normal secondary
f
throughput increases rapidly. When N increases further, the bandwidth occupied for
sending local decisions increases, which have more influence to the normal sec~ opt decreases instead. Consequently, it can be obtained that
ondary throughput, so C
not the larger N is certain to result the higher secondary throughput. Under different
sensing channels, in order to meet the target detection probability, the sensing
diversity order for maximizing the secondary throughput is different when using the
optimal algorithm. For example, when the average SNR of the sensing channels is
5 dB, the sensing diversity order is 3 for maximizing the secondary throughput, and
the maximized secondary throughput can be achieved only by three SUs cooperation when ensuring sufcient protection to the PU. When the average SNR of the
sensing channels is 10 dB, the sensing diversity order can be set as 2 for maximizing the secondary throughput, and the maximized secondary throughput can be
achieved only by two SUs cooperation when ensuring sufcient protection to the
PU. Consequently, when the average SNR of the sensing channels is given, it is
un-useful to increase the number of SUs for cooperation, so as not to induce the
unnecessary resources waste.
In CRN, because of the SUs mobility and the surroundings changing characteristics, the conditions of sensing channels are constantly changed. We can get the
number of SUs for maximizing the secondary throughput by the optimal algorithm
in different conditions of sensing channels, without selecting more SUs to participate in cooperation to save the system overheads.
Figure 3.12 shows the CSS performance using optimal algorithm with (Pe = 0.1)
and without (Pe = 0) reporting error respectively, in which, K = 1, SNR = 10 dB,
d 0:9. It is seen that the false alarm probability of CSS
P0 = 0.5, = 0.1, Q
monotonically decreased with N increase when there is no reporting error. However
when the reporting error exists, the false alarm probability of CSS decreases rst
and increase after. This is because it is less sensitive of P f to k, the influence of N to
Qopt
f larger than P f . The similar conclusions to normal secondary throughput could
be obtained in same theory.
(a)
65
0.9
pe=0
0.8
pe=0.1
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
10
the number of secondary user:N
(b) 0.7
pe=0,T-CSS
pe=0.1,T-CSS
0.6
0.5
0.4
0.3
0.2
0.1
0
1
10
Fig. 3.12 The performance of CSS under various reporting channels using the proposed method.
a Qf versus the number of sensing users. b The normal secondary throughput versus the number of
sensing users
66
3.5
Summary
In this chapter, we studied the CSS performance under both imperfect sensing
channels and the reporting channels when the local sensing results are parallelly
sent to the FC by OFDMA mode. Firstly, we analyzed the CSS performance based
on decision fusion. Secondly, we have made in-depth research on the relationship
among the number of sensing users, detection threshold, bandwidth, spectrum
utilization and the secondary throughput by considering both primary users and
secondary users interests. Particularly, we study the problem of designing the
detection threshold to maximize the achievable throughput of secondary users
under the constraint that the primary users are sufciently protected. Theoretical
analysis and simulation results show that the proposed optimal algorithm is effective in improving spectral efciency and achievable throughput of secondary users.
Meanwhile, it is flexible for different number of sensing users.
References
1. Federal Communications commission (2002) Spectrum policy task force, ET docket no.
02-135 Memorandum opinion and order
2. Akyildiz IF, Lee WY, Vuran MC et al (2006) Next generation/dynamic spectrum
access/cognitive radio wireless networks: a survey. Comput Netw J (Elsevier) 50(13):
21272159
cognitive radios. In: Proceeding of the 38th asilomar conference signals, system, computers.
5. Weiss T, Jondral F (2004) Spectrum pooling: an innovative strategy for the enhancement of
spectrum efciency. IEEE Commun Mag 42(3):S8S14
8. Ganesan G, Li Y (2007) Cooperative spectrum sensing in cognitive radio, part II: multiuser
environments. In: Proceedings of the 1st IEEE international symposium new frontiers dynamic
spectrum network DySPAN. Baltimore, MD, United states, pp 131136
11. Sun CH, Zhang W, Letaief KB (2007) Cluster-based cooperative spectrum sensing in
cognitive radio systems. In: IEEE international conference on communications, ICC 07.
Glasgow, pp 25112515
12. Zheng SK, Liang YC, Kam PY, et al (2009) Cross-layered design of spectrum sensing and
MAC for opportunistic spectrum access. In Proceedings 2009 IEEE wireless communications
and networking conference, WCNC 2009. Budapest, Hungary, pp 16
References
67
13. Zeng YH, Liang YC (2009) Spectrum-sensing algorithms for cognitive radio based on
statistical covariances. IEEE Trans Veh Technol 58(4):18041815
14. Yngve S, Tullberg H, Kronander J (2008) Sensor Selection for Cooperative Spectrum Sensing.
3rd IEEE symposium on new frontiers in dynamic spectrum access networks(DySPAN). IL,
United States, Chicago, pp 111
15. Bhargava VK (2007) Adaptive wireless access system design for cognitive radio networks.
2007 IEEE radio and wireless symposium. Long Beach, CA, pp 56
16. Lee W-Y, Akyildiz IF (2008) Optimal spectrum sensing framework for cognitive radio
18. Shen JY, Jiang T, Liu SY et al (2009) Maximum channel throughput via cooperative spectrum
sensing in cognitive radio networks. IEEE Trans Wireless Commun 8(10):51665175
19. Pei YY, Hoang AT, Liang YC (2007) Sensing-throughput tradeoff in cognitive radio
networks: how frequently should spectrum sensing be carried out? In: 18th annual IEEE
international symposium on personal, indoor and mobile radio communications, PIMRC07.
Greece, Athens, pp 18041805
20. Peh ECY, Liang YC, Guan YL et al (2009) Optimization of cooperative spectrum sensing in
cognitive radio networks: a sensing-throughput tradeoff view. IEEE Trans Veh Technol
58(9):52945299
Proc IEEE 97(5):878893
under bandwidth constraints. IEEE Wireless Commun. Networking Conference WCNC,
Kowloon, China, pp 15
23. Zhang W, Mallik R, Letaief KB (2009) Optimization of cooperative spectrum sensing with
energy detection in cognitive radio networks. IEEE Trans Wireless Commun 8(12):57615766
24. Zhang W, Mallik RK and Letaief KB (2008) Cooperative spectrum sensing optimization in
cognitive radio networks. In: IEEE international conference on communications. China,
pp 34113415
25. Digham FF, Alouini MS, Simon MK (2007) On the Energy detection of unknown signals over
fading channels. IEEE Trans Wireless Commun 55(1):2124
26. Urkowitz H (1967) Energy detection of unknown deterministic signals. Proc IEEE 55(4):
523531
in cognitive radio networks-[transaction letters]. IEEE Trans Wireless Commun 7(12):
47614766
28. Shen J, Liu S, Zeng L et al (2009) Optimisation of cooperative spectrum sensing in cognitive
radio network. IET Commun 3(7):11701178
29. Boyd S P. Vandenberghe L (2004) Convex optimization. Cambridge University Press, USA
30. Rifths DV, Smith IM (1991) Numerical methods for engineers. Blackwell Scientic, London
31. Zhao SJ, Zhao JX (2005) Signal detection and estimation theory. Tsinghua University Press,
Beijing
32. Li ML, Yuan CW, Li L et al (2011) Performance analysis and optimization of cooperative
spectrum sensing for maximizing secondary throughput. Tongxin Xuebao 32 (2):5360
33. Li ML, Yuan CW, Li L (2011) Analysis of secondary throughput and optimization in
cooperative spectrum sensing. J Chin Univ Post Telecom 18(4):3944
Chapter 4
The Censoring Based CSS
A censoring based CSS is illustrated in this chapter, which can be used to save the
system bandwidth and improve the system performance effectively. The analyze
model, performance (including detection performance and secondary throughput)
and some optimal algorithm are introduced for detail.
4.1
Introduction
It has been shown that there requires a control channel for each SU to report its
sensing result, which is usually bandwidth limited [1]. If every SU transmits the real
value of its sensing observation, innite bits are required, which will result in a
huge control bandwidth. Quantization of local observations has attracted much
research interest even though it introduces additional noise and a signal-to-noise
ratio (SNR) loss at the receiver [2]. A lot of work has been done on the quantization
for the signal detection in CSS, but most of them focused on the optimal design of
the quantizer [3, 4]. It was shown that two or three bits quantization was most
appropriate without noticeable loss in the performance [5]. It has been shown that
identical binary quantization, i.e., one bit quantization, performs asymptotically
optimal as the number of users goes to innity [6, 7]. However, when the number of
sensing users is very large, the total number of sensing bits transmitted to the FC is
still very huge.
Recently, censoring sensors have attracted a lot of attentions in decentralized
detection under communication constraints. In Ref. [8], the authors studied the
performance of censoring sensors under perfect reporting channels in the decentralized detection system, in which, only the likelihood ratios with enough information are allowed to send to the common receiver via perfect reporting channels.
In [9], although the imperfect reporting channels were taken into consideration in
censoring sensors, only the special case that the probability of the presence of the
primary user was sufciently small was investigated. The aforementioned
DOI 10.1007/978-3-319-31095-4_4
69
70
4 The Censoring Based CSS
researches assume orthogonal channels between the sensors and the FC, implying
large system bandwidth or large latency for large-scale scenario. In practice, it may
be challenging and wasteful to reserve an orthogonal channel for each sensor in the
network. Therefore, it may be more practical to assign nonorthogonal channels to
the sensors. In [10], a nonorthogonal signaling scheme for the transmission of
censored decisions and noncoherent fusion rules was proposed, in which, the
bandwidth efciency was achieved at the expense of only a small loss in power
efciency. All of the above researches are based on sensor networks. In [7], the
authors investigated the CSS performance under the limited bandwidth in CRN, in
which, only the users with reliable information are allowed to report their local
binary decisions (0 or 1) to the FC while the others will not make any decision
during the reporting stage. It has been shown that the average number of sensing
bits decreases greatly with a little loss of sensing performance by this censoring
method, and the control bandwidth occupied for sending local sensing results was
also reduced.
The above studies on censoring scheme show that: the control bandwidth used
for sensing local spectrum sensing results can be effectively saved by censoring
method. According to the researches in the second chapter, it is very important to
select the reasonable detection threshold, which will directly affect the detection
performance. However, the accurate detection threshold selection algorithm was not
given in all of the studies above.
Because in the decision fusion based CSS, the detection to 1 and 0 is relative.
As long as we know the probability of detection of 1 then we can get the probability of detection of 0. In order to save the control bandwidth occupied by
sending local sensing results and reduce the transmission for those no-meaningful
information, we proposed a new scheme of CSS based on censoring (C-CSS). We
found that not only the control bandwidth could be effectively saved by the C-CSS,
but also the advantages could be achieved that the CSS performance could not
depend on the number of sensing users, which could improve the CSS performance
under the imperfect reporting channels.
In this paper, we present the calculating method of the local decisions in C-CSS.
We also analyze the performance of C-CSS under both imperfect sensing channels
and reporting channels. An optimal detection algorithm is suggested further from
the point of secondary user. The analytical results show that the control bandwidth
can be effectively saved, the dependence of reporting error to the number of sensing
user can also be reduced by C-CSS. The tail sensing problem can be effectively
solved and the false alarm probability also decreased. The maximized spectrum
access chance of secondary user can be realized while assuring target detection
probability by optimal algorithm. In addition, the C-CSS performance from the
point of secondary user throughput is analyzed further. The relationship among SU,
detection threshold, bandwidth, spectrum utilization and secondary throughput is
studied and the optimal algorithm for maximizing the secondary throughput is
proposed. The research results indicate that the required detection threshold for
maximizing the secondary throughput is different between the traditional CSS and
4.1 Introduction
71
C-CSS under given SU number while assuring object detection probability. The
larger secondary throughput can be achieved with less SU in C-CSS. The optimal
secondary throughput can be realized by the proposed optimal algorithm.
4.2
System Model
The process of centralized based CSS in cognitive radio system is shown in

Fig. 4.1. The original status of the PU is assumed as H0 or H1, representing the
vacant spectrum present or not. SUi (i = 1, 2, N) use energy detection method to
detect the local spectrum independently, and send the decision results of Di (Di 2
{0, 1}) through the reporting channels to the FC. Here, the OFDMA mode is
considered in sending the local sensing results as detailed description in Chap. 3.
In the traditional CSS [11, 12], when ith SU performs local spectrum sensing,
whether the energy detected by ith SU is greater than a predened threshold i, the
SU will send the decision value to the FC. When the energy value is greater than
the threshold, it will send 1, on the contrary when the energy value is less than the
threshold, it will send 0. Obviously, in the above process, no matter the local
decision is 0 or 1, it will occupy a certain control channel bandwidth. In the CSS
based on decision fusion, the FC adds all the binary local decisions and then
executes a total decision. In this kind of decision fusion rule, decision value 0 has
actually no effect on the nal fusion results. Obviously, it will waste a lot of
bandwidth when sending local decision value in the traditional algorithm, especially
in the occasion that the probability of the PUs presence is small and the sensing
condition is good, where, the probability of the sensing decision results for 0 is
relatively large which will occupy too much control channel bandwidth.
In order to save resources, we propose a new censoring based CSS scheme
(C-CSS) in this paper, as shown in Fig. 4.2 [13, 14]. When the ith SU execute the
local spectrum sensing, only when the detected energy value of ith SU is lager than
the predened threshold i, i.e. the local decision value of ith SU is 1, the local
decision value will be sent to the FC, otherwise ith SU will send nothing and
continue to detect the PU signal or remain silent. Obviously, the amount of data
sent to the FC can be greatly reduced by using C-CSS scheme, which results the
reduced complexity for the FC. In the limited bandwidth system, when each SU
concurrently send its decision value to the FC in different channel, more control
Fig. 4.1 The CSS process

under imperfect reporting
channels
SU1
PU
H0/H1
SU2
SUN
D1
Reporting
channel 1
D2
Reporting
channel 2
DN
Reporting
channel N
FC
72
Fig. 4.2 The sending mode

for local spectrum sensing
results in the C-CSS.
a Traditional CSS. b C-CSS
(a)
send
send 1
(b)
0
Yi
send
i
Yi
channel bandwidth can be saved for data transmission by C-CSS scheme, which is
especially meaningful in the process for signaling interaction of MAC layer.
If we dene the set 1 composed by all the decision result of 1 and 0
composed by all the decision result of 0, the SU belongings to 0 can be directly
regarded as a relay node to assist other nodes for transmitting data in cooperative
communication, rather than through three times handshake protocol to save the
signaling overhead. Of course, it will lost some information, but in the high density
network with poor sensing channel conditions, the error detection probability of the
PU signal is relatively large, then the proposed method will have important realistic
signicance.
4.3
4.3.1
The C-CSS Detection Performance

Performance Analysis
According to the introduction to the C-CSS scheme in the above section, we can get
the C-CSS process steps as follows:
(1) The ith SU performs local spectrum sensing independently using energy
detection method, the detected energy value is expressed as Yi , if the energy
value is greater than a predened threshold value ki , ith SU makes a binary
decision of Di , otherwise it will do not perform a local decision, then, the
decision rules can be expressed as:
Di 1;
Yi [ ki
4:1
Let MC represents the actual number of local decision in the C-CSS, and we
assume that the signal received by each SU is independent and identically distributed. According to the model described above, based on the Bayes theory, the
mean value of MC can be calculated as follows (4.2):
MC dN P0 PfY [ kjH0 g P1 PfY [ kjH1 ge
4:2
4.3 The C-CSS Detection Performance
73
where, P0 indicates the probability of the free channel which is not occupied by the
PU, P1 indicates the probability of the occupied channel, P0 + P1 = 1; de denotes
taking upper integer.
(2) The ith SU send the decision results to the fusion center through reporting
channel.
We consider the channels between the PU and SUs (i.e. sensing channels) are
Rayleigh fading channels in this chapter. The calculation of the local false alarm
probability of Pf and the detection probability of Pd have been described in Chap. 3.
If the SUs send 0 while the FC judges as 1, we tag this probability of error
transmission from 0 to 1 as qf . If the SUs send 1 while the FC judges as 0, we
tag this probability of error transmission from 1 to 0 as qm .
In the traditional CSS, let P f denotes SUs false alarm probability, P d denotes
SUs detection probability before the FCs fusion, and Qf is the total false alarm
probability, Qd is the total detection probability of CSS, according to the analysis in
Chap. 3, their expressions can be described as follows:

P f Pf 1 qm 1 Pf qf
4:3
P d Pd 1 qm 1 ^pd qf
4:4
Qf
N
X

B j; N; P f ; 1 P f
4:5
jK
Qd
N
X
Bj; N; P d ; 1 P d
4:6
jK
p j qnj .
j
Because only decision value 1 will be sent to the FC in the C-CSS scheme, the
false alarm probability P f ;CCSS and the detection probability P d;CCSS before
fusion can be described as:
where, K 2 f1; 2; . . .; N g; Bj; n; p; q
P f ;CCSS Pf 1 qm
4:7
P d;CCSS Pd 1 qm
4:8
(3) The FC fuses all the received decisions.

In the C-CSS scheme, only when the local detection result is 1, the SU will
send the local decision value 1 to the FC through reporting channel. Combining
Eqs. (4.7) and (4.8), when the FC use k out of n fusion rule to implement fusion,
74
the nal false alarm probability Qf ;CCSS and the detection probability Qd;CCSS of
the C-CSS can be respectively expressed as:
Qf ;CCSS
MC
X

B j; MC ; P f ;CCSS ; 1 P f ;CCSS
4:9
jL
Qd;CCSS
MC
X

B j; MC ; P d;CCSS ; 1 P d;CCSS
4:10
jL
where, K 2 f1; 2; . . .; MC g.
The calculation to the detection performance of the traditional CSS and C-CSS
has been given above. We will analyze the performance in theory in the following.
By comparing Eqs. (4.3) and (4.7), it can be seen that, the false alarm probability
of P f ;CCSS in the C-CSS scheme is less than the false alarm probability of P f in the
traditional CSS scheme, which are both influenced by qf and qm . Because qf and qm
are both the error transmission probability in reporting channels, according to the
symmetric characteristic of channel, let qf qm pe , then, Eqs. (4.3) and (4.7)
change into:

P f Pf 1 Pe 1 Pf Pe Pf 1 2Pf Pe
4:11
P f ;CCSS Pf 1 Pe Pf Pf Pe
4:12
From Eqs. (4.11) and (4.12), we can see that the false alarm probability of P f in
the traditional CSS scheme equals to the addition of Pf and a positive value which
varies with the report error probability of Pe ; the false alarm probability of P f ;CCSS
in the C-CSS scheme equals to the subtraction of Pf with a positive value which
varies with the report error probability of Pe . Obviously, the expression P f ;CCSS
P f Pf is valid when Pe 0, otherwise, P f ;CCSS \P f . The advantage of C-CSS
scheme changes with Pe can be obtained by the subtraction of Eq. (4.11) with
Eq. (4.12), i.e. e P f P f ;CCSS 1 Pf Pe . It is seen that the larger the value
of Pe the larger the value of , i.e. the advantage of the C-CSS scheme is more
obvious, which is mainly due to the fact that the false alarm events are reduced in
the reporting channels, i.e. avoid the occasion that the FC assume 1 and SU
send 0.
The number of the local decisions which have been transmitted to the FC can be
expressed as
MC dN P0 PfY [ kjH0 g P1 PfY [ kjH1 ge
dN P0 1 PfY\kjH0 g P1 1 PfY\kjH1 ge
dN 1 P0 PfY\kjH0 g P1 PfY\kjH1 ge N
4:13
75
Table 4.1 The false alarm probability comparison between the C-CSS and the traditional CSS
under different Pe
Pe
0.001
0.01
0.1
The false alarm probability

P f ; CCSS
Pf
Qf
Qf,
0.011
0.0198
0.108
0.0218
0.0392
0.2043
0.01
0.0099
0.009
0.01
0.0099
0.009
C-CSS
Compare Eqs. (4.5) and (4.9), because MC N in one hand, in the other hand the
inequality of P f ;CCSS P f can be easily obtained, then the inequality of
Qf ;CCSS Qf can be get.
In order to illustrate the problem more clearly, the intuitional comparisons
between the C-CSS scheme and the traditional CSS scheme under different Pe are
given in Table 4.1, in which, the parameters are set by Pf 102 , N = 2, MC = 1.
From the table we can see that, the value of P f ;CCSS is almost the same as the value
of P f when Pe is relatively small. The advantage of the optimal algorithm becomes
more obvious with the growing of Pe , i.e. P f ;CCSS \P f . It also can be seen that the
total false alarm probability can always be decreased signicantly by the optimal
algorithm no matter how Pe is, i.e. Qf ;CCSS \Qf .
When we use the k out of n fusion rule, the overall performance trends will not
be affected by the value of k. Therefore, in order to simply illustrate the influence of
sensing performance by the reporting fading channels, we take k = 1, i.e. L = k = 1
as example to analyze, and we assume that each SU is independent and identically
distributed. In the following, we will prove that the serious tail sensing problem
exists in the traditional CSS when the reporting channels suffer from fadings, which
however, can be effectively solved by the C-CSS scheme.
Proof According to Eq. (4.5), when K = 1, we have

N

N
Qf 1 1 P f 1 1 Pe 1 2Pe Pf
4:14

N1
@Qf
N 1 Pe 1 2Pe Pf
1 2Pe
@Pf

N1
N 1 Pe 1 2Pe Pf
1 2Pe
4:15
Considering the reporting probability in practice is not more than 1/2, there are:
@Qf
0
@Pf
4:16
We can see that Qf is monotonically increased with Pf , so there exists minimum

value of Qf . And because Pf 0, the minimum value of Qf named as Qf ;min reaches
minimal at the spot of Pf 0, that is:
76
Qf ;minjPf 0 1 1 Pe N NPe
4:17
According to Eq. (4.17), we can see that there exists a false alarm lower bound,
i.e. the minimum false alarm probability in the traditional CSS scheme. When the
reporting error probability of Pe is not equal to zero, the lower bound increases with
the increase of Pe and the number of sensing users. The following simulation results
also show that, from the detection performance curve, the detection probability
decreases drastically at the lower bound Qf ;minjPf 0 . Although the detection performance can be improved by multiple SUs cooperative spectrum sensing in
theory, the performance will be seriously affected by the error transmission in the
reporting channels.
As for the C-CSS scheme, from Eq. (4.9), we can get that:

M
Qf ;CCSS 1 1 1 Pe Pf c
4:18

M 1
@Qf ;CCSS
MC 1 1 Pe Pf c 1 Pe
@Pf

M 1
MC 1 1 Pe Pf c 1 Pe [ 0
4:19
Thus, the minimum value of Qf ;CCSS is

Qf ;CCSS;minjPf 0 0
4:20
It can be seen that the minimum value of Qf ;CCSS is irrespective with the
reporting error probability and sensing users, i.e. the C-CSS performance will not
be impacted by the reporting error probability, which have effectively solved the tail
sensing problem and can perform the cooperative spectrum sensing better.
4.3.2
Optimal Algorithm
According to the above analysis, the lower false alarm probability can be obtained
by the C-CSS scheme compared with the traditional CSS under the same detection
threshold. However, in practice, what we concern more about is how to increase the
access opportunity of SU when assuring sufcient protection to the PU, i.e. how to
minimize the false alarm probability when ensuring that the target detection
probability of is enough large. In this paper, we establish an optimal model by
using the N-P criterion as follows:
Minimize Qf
Subject to Qd g
4:21
77
In Chap. 3, we have proved the following two Summary:

Conclusion 1: the false alarm probability of Qf and the detection probability of Qd
are both monotonically decreased with the detection threshold , that is
@Qf
\0
@k
4:22
@Qd
\0
@k
4:23
Conclusion 2: the false alarm probability of Qf and the detection probability of Qd

are both monotonically increased with sensing diversity order represented by N,
that is
@Qf
[0
@N
4:24
@Qd
[0
@N
4:25
From conclusion 1 and conclusion 2 we can see that, when the number of
sensing diversity order is given, the higher the detection threshold is, the lower the
false alarm probability Qf , and more chances for SUs to access, but the detection
probability may decrease which will create interference to the PU. On the other
hand, when the detection threshold is given, the higher sensing diversity order will
make the higher detection probability and the higher false alarm probability, which
will make fewer chances for SUs to access. Therefore, there exists a tradeoff
between the sensing diversity order and the detection threshold.
On the other hand, comparing to the traditional CSS scheme, the number of local
decisions transmitted to the FC decreases in the C-CSS scheme, in which, the
sensing diversity order does not equal to the number of sensing users. In order to
guarantee the target detection probability denoted by , we have to decrease the
detection threshold () when the number of local decisions transmitted to the FC
(M) decreases. However, the value of the false alarm probability Qf will decreases
with the decrease of M and increases with the decrease of , which makes it difcult
to judge whether the false alarm probability will be increase or not. The simulation
results show that the lower false alarm probability can be obtained by the C-CSS
scheme compared to the traditional CSS scheme when the detection threshold is
given. When satisfying the target detection probability of = 0.9, the lower
detection threshold is needed in the C-CSS scheme.
78
In order to obtain the optimal detection performance under different SU and to

increase the access opportunity for SU, we give the realization process of the
optimal algorithm for minimizing the false alarm probability as follows. When the
number of sensing users (N) is given, we select a detection threshold k0 make
Qd k0 g, then select a threshold for k1 k0 , we have Qd k1 Qd k0 ,
Qf k1 Qf k0 . In order to ensure the PU is adequately protected and will not be
disturbed, the optimal solution to Eq. (3.21) can be achieved with equality constraint in Eq. (3.21), i.e. the maximized false alarm probability can be achieved in
maintaining PU sufciently protected when k0 is selected as detection threshold.
Therefore, the optimization objective changes into how to nd the optimal detection
threshold value of k0 .
Firstly, the optimal algorithm for traditional CSS is given as follows:
(1) According to the analysis above and the optimal algorithm in Chap. 2, we can
see that in the traditional CSS, the explicit expression of the detection probability by the detection threshold exists. Therefore, for a certain N, the optimal
detection threshold of kopt
N can be obtained by solving the equation of Qd k
g by Newton-Raphson method. When using the Newton-Raphson method to
get the optimal value, we should set objective optimization function rstly, let
gkN Qd kN g, i.e. let the difference error between the actual detection
probability and the target detection probability minimized, then implement as
the following steps:
Step 1 choose tolerance and the initial guess kN 1, let i = 1;
Step 2 if jgkN ij\e, stop; otherwise, go to step 3;
Step 3 let kN i 1 kN i gkN i=g0 kN i. Then, let i = i + 1 and go to
step 2.
Through the above three steps, we can get the optimal detection threshold for
minimizing the false alarm probability when the target detection probability is
satised. It is needed to note that we should combine Eqs. (4.4) and (4.6) to
get the above solution.
opt
(2) Calculate the corresponding local false alarm probability Popt
f kN under
Rayleigh fading sensing channels.
opt
(3) Calculate the corresponding total false alarm probability of CSS Qopt
f kN
according to the Eq. (4.5).
The above is the realization process for the optimal false alarm probability in the
traditional CSS. Due to the actual local decisions transmitted to the FC are different
in the C-CSS scheme and the traditional CSS, and the number of MC in the C-CSS
scheme is related to the detection threshold. Therefore, the optimal detection
threshold can not be calculated directly by using the existing optimal algorithm in
the C-CSS scheme. Since the essential of the optimal algorithm is actually a kind of
transformation of the original iterative method, the different iterative method makes
79
the different algorithm precision and complexity, we can obtain the optimal
detection threshold for minimizing the false alarm probability by utilizing the
essential of the optimal algorithm in the C-CSS scheme. In the following, we
exploit the two methods of iterative method and the successive approximation
method.
For the C-CSS scheme, the actual local decisions transmitted to the FC are
different in the C-CSS scheme and the traditional CSS, however, both of them are
related to the number of sensing users N. What we concern in this paper are the
relationships among the number of sensing users N and the detection threshold
and other performance parameters. Consequently, the progresses of the optimal
algorithm for C-CSS scheme can be processed as follows in this section.
(1) For a certain N, calculate all of the actual local decisions transmitted to the FC
(MC) and the detection probability corresponding to all the detection
threshold

in the C-CSS scheme, the vector representation is: Qd;CCSS Qd;CCSS k1 ;
Qd;CCSS k2 ; . . .; Qd;CCSS kn g.
(2) Let D Qd;CCSS g I fvk1 ; vk2 ; . . .; vkn g, Where I is the n-dimensional unit vector.
(3) Choose
tolerance
n 104 , and let j = 1.

(4) If vkj \n, stop, and tag the corresponding j, otherwise, let j = j + 1,
continue.
Through the above steps, we can get the optimal threshold detection
kopt
N;CCSS for maximizing the access opportunity of SU when there exists
different SU number in the C-CSS scheme.
opt
(5) Calculate the local false alarm probability Popt
f ;CCSS kN;CCSS corresponding to
opt
the optimal detection threshold value kN;CCSS in the C-CSS scheme.
(6) Using the results in step (5), calculate the total false alarm probability
opt
opt
Qopt
f ;CCSS kN;CCSS corresponding to the optimal detection threshold kN;CCSS
according to Eq. (4.9).
4.3.3
Simulation Results
Figure 4.3 illustrates how the bandwidth is saved by the C-CSS scheme under
different SNR of sensing channels and different probabilities of idle channels
denoted by P0 respectively. The abscissa is the local detection threshold of
(normalized value), the ordinate is the ratio of the actual local decisions transmitted
to the FC to the number of sensing users denoted by MC/N. From Fig. 4.3, we can
see that with the increase of the detection threshold , the value of MC/N reduces. In
fact, a higher will induce the decrease of the false alarm probability Pf and the
detection probability Pd , which will result the fewer number of the local decision
number of 1 and the decrease of MC, correspondingly, the occupied control
80
1
the ratio of the actual trnasmitted bits to

the number of sensing users
Fig. 4.3 The bandwidth

saved by the C-CSS scheme
SNR=0dB,p0=1/3
SNR=0dB,p0=2/3
SNR=5dB,p0=1/3
0.9
SNR=5dB,p0=2/3
SNR=10dB,p0=1/3
SNR=10dB,p0=2/3
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
10
15
20
25
30
the detection threshold
bandwidth will be reduced. From Fig. 4.3, it also can be seen that for a given
detection threshold , with the increase of the average SNR in sensing channels, the
control bandwidth occupied for sending local decisions will be increased, which is
mainly because the better the sensing channel conditions, the larger the number of
the local decisions of 1, which means that more control bandwidth are needed to
send the local decisions. A larger idle channel probability of P0 results the less need
of the control bandwidth for sending the local decisions. This is mainly because the
actual PU being operated is relatively less, and there are more local decisions of 0,
which makes fewer control bandwidth occupied.
Figure 4.4 shows the detection performance varied with the number of sensing
users N in the traditional CSS (T-CSS) and C-CSS scheme respectively. The
simulation parameters are set by P0 = 0.5, Pe = 0.1, SNR = 10 dB, u = 5. It can be
seen that the false alarm probability can be reduced by the C-CSS scheme. On the
-0.1
10
the detection probability
Fig. 4.4 The detection

probability versus the false
alarm probability under
different sensing users of
N with different schemes
-0.2
10
-0.3
10
N=1,T-CSS
N=2,T-CSS
N=4,T-CSS
N=10,T-CSS
N=1,C-CSS
N=2,C-CSS
N=4,C-CSS
N=10,C-CSS
-0.4
10
-0.5
10
-0.6
10
-2
10
-1
10
10
-0.1
the detection probability
Fig. 4.5 The detection

probability versus the false
alarm probability under
different Pe with different
schemes
81
10
-0.2
10
-0.3
10
pe=0.001,T-CSS
pe=0.01,T-CSS
pe=0.1,T-CSS
pe=0.001,C-CSS
pe=0.01,C-CSS
pe=0.1,C-CSS
-0.4
10
-3
10
-2
10
-1
10
10
other hand, the lower bound of the false alarm probability exists in the T-CSS
scheme, i.e. at this spot, the detection probability will be drastically varied, and
which is increased with the increase of sensing users. However, the lower bound of
the false alarm probability does not exist in the C-CSS scheme, which effectively
solve the tail sensing problems existed in the T-CSS scheme, and realize that the
greater the number of sensing users the better the sensing performance, with is
consistent with the theoretical analysis.
Figure 4.5 illustrates the detection performance comparison between T-CSS and
C-CSS under different reporting error probabilities. The simulation parameters are
set by P0 = 0.5, N = 2, SNR = 10 dB, u = 5. It is seen that the false alarm bound
increases with the increase of Pe . The overall performance trend of the C-CSS
scheme is nearly not affected by Pe , and the performance advantages of the C-CSS
scheme become obvious with the increase of Pe . For example, when Pe 0:1, not
only the false alarm probability can be reduced but also the detection probability
can be improved by the C-CSS scheme.
Figure 4.6 shows the performance comparison between T-CSS and C-CSS under
different SNR of sensing channels after utilizing the optimal algorithm when
assuring the target detection probability of 0.9, in which, Pe 0:1, u = 5. It can be
seen that the lower false alarm probability can be obtained by C-CSS compared to
T-CSS, which decreases with the increase of N in the C-CSS scheme but decreases
rst and then increases in the T-CSS scheme. Therefore, when considering the
practical applications, the access opportunities for SUs can be increased by the
C-CSS scheme. In addition, for a certain N, the false alarm probability will be
reduced with the increase of the average SNR in sensing channels. When the
sensing channel conditions is relatively poor such as SNR = 0 dB, it is not obvious
for improving the detection performance by the C-CSS scheme, which is mainly
due to the poor local detection performance, consequently, the optimal false alarm
probability is nearly the same for T-CSS and C-CSS.
82
0.9
0.8
Fig. 4.6 The false alarm

probability versus sensing
users under different SNR
by optimal algorithm
0.7
0.6
0.5
0.4
0.3
SNR=0dB,T-CSS
SNR=5dB,T-CSS
0.2
SNR=10dB,T-CSS
SNR=0dB,C-CSS
0.1
SNR=5dB,C-CSS
SNR=10dB,C-CSS
10
Figure 4.7 shows the performance comparison between T-CSS and C-CSS with
reporting error and without reporting error after utilizing the optimal algorithm
when assuring the target detection probability of 0.9, in which, SNR = 10 dB, u = 5.
When reporting error exists, the lower false alarm probability can be achieved by
C-CSS compared to T-CSS. When there does not exist reporting error, the lower
false alarm probability can be achieved by T-CSS compared to C-CSS, which is
because more sensing users are needed to assure the target detection probability in
C-CSS when Pe 0. It is needed to note that there exist some large fluctuations for
the false alarm probability in Figs. 4.6 and 4.7, which is mainly because that MC
must be an integer, on the other hand, in order to make sure sufcient protection to
the PU, we utilize the upper integral method.
0.9
pe=0,T-CSS
pe=0.1,T-CSS
0.8
pe=0,C-CSS
Fig. 4.7 The false alarm

probability versus sensing
users under different Pe by
optimal algorithm
pe=0.1,C-CSS
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
the number of secondary user:N
10
4.4
83
The C-CSS Secondary Throughput
4.4.1
4.4.1.1
Spectrum Utilization
According to the analysis in Chap. 2, the percentage of the correctly identied

vacant spectrum provides a reasonable measure of the spectrum utilization, which is
determined by 1 Qf . The smaller Qf leads to the higher percentage of correctly
identied vacant spectrum, which resulting the improved spectrum utilization. Qf is
related to the local detection threshold and the actual local decisions of
M transmitted by SUs. When is given, namely Pf is given, Qf is monotonically
increased with M. Consequently, a smaller M makes a lower Qf and improved
spectrum utilization. When N is given, a higher detection threshold makes the lower
false alarm probability of Pf and Qf and the higher spectrum utilization, whereas
the detection probability decreases which may cause interference to the PU.
Since the actual local decisions transmitted in C-CSS is less than that in T-CSS,
i.e. MC N. Consequently, under the same detection threshold, the total false alarm
probability of C-CSS is less than that of T-CSS, and thus the higher spectrum
utilization can be achieved in the C-CSS scheme.
4.4.1.2
The Secondary Throughput
In CRN, there exist two scenarios for SUs to access the channels: (1) the channel is
idle and is correctly detected by a SU; (2) the channel is occupied by a PU but is not
detected by SUs. In the second chapter, we analyzed the secondary throughput from
the point of bandwidth in CRN, and showed the calculation method of the normalized expressions. In the same theory, the expressions of the secondary
throughput in the C-CSS scheme can be calculated as follows

~ CCSS k 1 Ma 1 Qf ;CCSS k
C
4:26
where, = B/W denotes the ratio of the control bandwidth occupied for sending 1bit
local decision to the total bandwidth, M is the actual local decisions sending to the
FC.
It is seen that the secondary throughput is affected by both M and Qf . In one
hand, the smaller M, the less control bandwidth occupied for reporting the local
sensing results and the larger available bandwidth for SUs transmission, which will
make the secondary throughput increased. In the T-CSS scheme, the number of
actual local decisions sending to the FC is the number of sensing users (N) in the
network. Since the inequality of MC N exists, the number of actual local decisions
sending to the FC can be reduced by the C-CSS scheme, and the control bandwidth
saved, which makes the reduced false alarm probability and the improved
84
secondary throughput. On the other hand, in order to meet the target detection
probability, the local detection probability of Pd;CCSS Pd is required, then, the
required detection threshold of C-CSS is lower than that of T-CSS, which results
the increase of the local false alarm probability Pf ;CCSS in C-CSS. Consequently,
whether the nal value of Qf ;CCSS increases or not can be determined directly, and
thus the secondary throughput how to change is not sure. The following simulation
results show that, for a given , the lower false alarm probability and the higher
secondary throughput can be achieved by the C-CSS scheme compared to the
d 0:9, the
T-CSS scheme. When satisfying the target detection probability of Q
required detection threshold is lower in the C-CSS scheme, but the secondary
throughput will not reduce, i.e. compared to the T-CSS scheme, the higher secondary throughput can be obtained by using less sensing users in the C-CSS
scheme.
4.4.2
Optimal Algorithm
4.4.2.1
Optimal Model
According to the above analysis, in one hand, the number of actual local decisions
sending to the FC reduced in the C-CSS scheme compared to that in the T-CSS
scheme, which results the different secondary throughput under the same detection
threshold in the two schemes. On the other hand, under the premise of meeting the
same target detection probability, the required sensing users and the corresponding
detection thresholds are both different for maximizing the secondary throughput
between the two schemes. In the C-CSS scheme, the number of actual local
decisions sending to the FC of M and the detection threshold of are related with
each other, i.e. there exists an optimal pair of M; k to maximize the secondary
throughput when assuring the target detection probability.
~ CCSS is the function of . Similar to the
From Eq. (4.26) we can see that, C
Sect. 4.3.2, according to the N-P criterion, the optimal problem described above can
be transformed into the following constraints:
(

~ CCSS k 1 Ma 1 Qf ;CCSS k
max C
k
d
s:t: Qd;CCSS Q
4:27
We have illustrated the optimal algorithm realization process for maximizing the
secondary throughput in the T-CSS scheme in Sect. 3.4. In the T-CSS scheme, the
number of actual local decisions sending to the FC (M) is equal to the number of
sensing users (N) in the network. Therefore, the optimal detection threshold for
maximizing the secondary throughput is the solution for satisfying the constraints
d . While in the C-CSS scheme, when the sensing users are certain in
of Qd k0 Q
d,
the network, we select a detection threshold of kC0 , makes that Qd;CCSS kC0 Q
85
let M0 denotes the number of actual local decisions sending to the FC corresponding
to the detection threshold of kC0 , and we select an arbitrary detection threshold make
that kC1 kC0 , let M1 denotes the number of actual local decisions sending to the FC
corresponding to the detection threshold of kC1 . The following equations may
not exist any more since M1 [ M0 : Qd;CCSS kC1 Qd;CCSS kC0 , Qf ;CCSS kC1
~ CCSS kC C
~ CCSS kC .
Qf ;CCSS kC0 , C
1
0
Consequently, in the C-CSS scheme, the optimal detection threshold for maximizing the secondary throughput can not be obtained by solving the equation of
d.
Qd;CCSS kC0 Q
4.4.2.2
Algorithm Design
Although the number of actual local decisions sending to the FC are different in the
C-CSS scheme and the T-CSS scheme, both of them are related to N. What we
concerned is the relationships among N, the detection threshold and other performance parameters. The optimal objective is to maximize the secondary throughput
~ CCSS , which in essence is related to N and the detection threshold. Then, the
of C
optimal algorithm in this section can be described as follows:
(1) For the C-CSS scheme, since MC is related to the detection threshold , the
optimal value can not be directly obtained by using the existing algorithms. In
this section we will nd out the optimal solution for Eqs. (4.27) by using the
idea in Sect. 4.3.
Step 1: For an arbitrary number of sensing users of N in the network, nd
out all of the false alarm probabilities and the number of actual local
decisions sending to the FC corresponding to all of the detection
thresholds when satisfying the target detection probability in the
C-CSS scheme. And then, all the normalized the average secondary
throughput can be obtained when satisfying the target detection
~ CCSS
probability, vector representation of which denote as C

~ CCSS k1 ; C
~ CCSS k2 ; . . .; C
~ CCSS kN ;
C
Step 2: Dene the accuracy , and let i = 1, j = 1;
~ CCSS ki , let j = j + 1, continue this step, and stop
~ CCSS ki C
Step 3: If C
when the expression of j = N is satised; otherwise let i = i + 1,
continue.
Through the above steps we can get the optimal detection threshold of
kopt
N;CCSS for maximizing the secondary throughput when satisfying the target
detection probability corresponding to an arbitrary N in the C-CSS scheme.
86
opt
(2) Calculating the local probability of false alarm Popt
f ;CCSS kN;CCSS and the

opt
probability of detection Popt
corresponding to the optimal
d;CCSS kN;CCSS
detection threshold kopt

N;CCSS .

opt
(3) Calculating the total false alarm probability of C-CSS Qopt
f ;CCSS kN;CCSS
corresponding to the optimal detection threshold kopt
N;CCSS according to
Eq. (4.9).

opt
~ opt
(4) Calculating the optimal normal secondary throughput C
k
CCSS
N;CCSS
according to Eq. (4.26).
From the processes of nding out the optimal value in the C-CSS scheme and the
T-CSS scheme, we can see that MC N, and in order to meet the target detection
probability, it is required that the local detection probability of C-CSS should be not
opt
lower than that of T-CSS, i.e. Popt
d;CCSS Pd . Therefore, the optimal detection
threshold for maximizing the secondary throughput in the C-CSS scheme may be
opt
less than that of in the T-CSS scheme, i.e. kopt
N;CCSS kN , which can be validated
by the following simulation results.
4.4.3
Simulation Results
4.4.3.1
Description
In this section, the simulation results mainly consist of two parts. The basic
simulation results are shown in the rst part, with the purpose of verifying the
difference of the spectral efciency and the secondary throughput in the C-CSS and
T-CSS scheme and then validating the reasonability of the proposed optimal
algorithm above as shown in Fig. 4.8. While in the second part, the performance
comparisons between C-CSS and T-CSS are illustrated after using the optimal
algorithm, which are used to verify the advantages of the C-CSS scheme and to
show that the higher secondary throughput with less SU can be achieved by the
C-CSS scheme. Because the number of actual local decisions sending to the FC is
related to the channel idle probability and the average SNR in sensing channels in
C-CSS, and the false alarm probability is related to the the average SNR in sensing
channels and the reporting error probability, we also give the performance curves
influenced by these parameters of C-CSS and T-CSS, including Figs. 4.9, 4.10
and 4.11.
(a)
0.9
0.8
0.7
0.6
0.5
N=1,T-CSS
0.4
N=4,T-CSS
0.3
N=10,T-CSS
N=2,T-CSS
N=8,T-CSS
N=1,C-CSS
0.2
N=2,C-CSS
N=4,C-CSS
0.1
N=8,C-CSS
N=10,C-CSS
10
15
20
25
30
20
25
30
Threshold
Normal throughput of second system
(b) 0.9
N=1,T-CSS
N=2,T-CSS
0.8
N=4,T-CSS
N=8,T-CSS
0.7
N=10,T-CSS
N=1,C-CSS
0.6
N=2,C-CSS
0.5
N=4,C-CSS
0.4
N=10,C-CSS
N=8,C-CSS
0.3
0.2
0.1
0
0
10
15
Threshold
(c)
1
0.9
Probability of detection
Fig. 4.8 The performance

comparison between the
C-CSS and the T-CSS
scheme. a The false alarm
probability versus the
detection threshold, b the
normalized secondary
throughput versus the
detection threshold, c the
detection probability versus
the detection threshold
87
0.8
0.7
N=1,T-CSS
N=2,T-CSS
N=4,T-CSS
0.6
N=8,T-CSS
N=10,T-CSS
0.5
N=1,C-CSS
N=2,C-CSS
N=4,C-CSS
0.4
N=8,C-CSS
N=10,C-CSS
10
15
Threshold
20
25
30
88
(a) 0.7

C-CSS and the T-CSS scheme
under different SNR by using
the optimal algorithm. a The
throughput versus the number
of sensing users, b the false
alarm probability versus the
number of sensing users
0.6
SNR=0dB,T-CSS
SNR=5dB,T-CSS
SNR=10dB,T-CSS
0.5
SNR=0dB,C-CSS
SNR=5dB,C-CSS
SNR=10dB,C-CSS
0.4
0.3
0.2
0.1
0
10
10
The number of secondary user: N
(b) 0.9
0.8
0.7
0.6
0.5
0.4
0.3
SNR=0dB,T-CSS
SNR=5dB,T-CSS
0.2
SNR=10dB,T-CSS
SNR=0dB,C-CSS
0.1
SNR=5dB,C-CSS
SNR=10dB,C-CSS
4.4.3.2
Results and Analysis
Figure 4.8 shows the detection probability, the false alarm probability and the
secondary throughput varied with the detection threshold under the different
number of sensing users N in C-CSS and T-CSS scheme. In the simulation, let
K = 1, Pe = 0.1, P0 = 0.5, SNR = 10 dB, = 0.1. From Fig. 4.8a, b, we can see that,
when the detection threshold is certain, the false alarm probability can be reduced
by the C-CSS scheme, and the secondary throughput can be greatly improved,
which become obvious with the increase of N. Especially when N = 10, i.e. the
bandwidth occupied for sending local decisions reaches the limit in the T-CSS
scheme which makes the secondary throughput as 0, however, in the C-CSS
scheme, the number of actual local decisions sending to the FC is less than 10
which makes the secondary throughput improved obviously. On the other hand, for
a given N, the false alarm probability decreases and the secondary throughput
(a) 0.7

under different P0 by using
the optimal algorithm. a The
89
0.6
0.5
0.4
0.3
Po=1/3,T-CSS
0.2
Po=1/2,T-CSS
Po=2/3,T-CSS
0.1
Po=1/3,C-CSS
Po=1/2,C-CSS
Po=2/3,C-CSS
10
(b) 0.9
0.8
0.7
0.6
0.5
Po=1/3,T-CSS
Po=1/2,T-CSS
0.4
Po=2/3,T-CSS
Po=1/3,C-CSS
Po=1/2,C-CSS
0.3
Po=2/3,C-CSS
0.2
0.1
0
10
increase with the increase of the detection threshold. While from Fig. 4.8c, we can
see that the detection probability decreases with the increase of the detection
threshold. The required detection threshold becomes lower in the C-CSS scheme
than that of in the T-CSS scheme when reaching the target detection probability of
d 0:9, but the secondary throughput will not be reduced thereafter, as shown in
Q
Table 4.2 for comparison. In conclusion, in order to meet the target detection
probability, the lower detection threshold is required in the C-CSS scheme than that
of in the T-CSS scheme under the same N. Compared to the T-CSS scheme, the
higher secondary throughput can be obtained by using less sensing users in the
C-CSS scheme to save the control bandwidth.
Figure 4.9 illustrates the performance comparison between the C-CSS and the
T-CSS scheme with different average SNR in sensing channels by using the optimal
algorithm. The normalized secondary throughput and the false alarm probability
varied with the number of sensing users N are shown in Fig. 4.9a, b respectively.
90
(a) 0.8

under different Pe by using the
optimal algorithm. a The
0.7
0.6
0.5
0.4
0.3
0.2
pe=0,T-CSS
pe=0.1,T-CSS
0.1
pe=0,C-CSS
pe=0.1,C-CSS
10
(b) 0.9
pe=0,T-CSS
pe=0.1,T-CSS
0.8
pe=0,C-CSS
pe=0.1,C-CSS
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
10
SU
Table 4.2 The comparison of the optimal detection threshold and corresponding secondary
throughput in the C-CSS scheme and the T-CSS scheme
Scheme
The
number
of SU (N)
The optimal detection threshold

for satisfying the target detection
probability ()
The normalized secondary

throughput corresponding to the
optimal detection threshold
T-CSS
2
4
2
4
15
27
13
14
0.5
0.38
0.52
0.9
C-CSS
From Fig. 4.9a, when N is xed, the secondary throughput increases with the
increase of SNR. This is mainly because in order to satisfy the target detection
probability, the required detection threshold becomes higher with the increase of
91
SNR, and the secondary throughput increases when the target false alarm probaopt
~ opt and C
~ opt ,
bility Qopt
CCSS
f and Qf ;C becomes smaller, which results the increase of C
~ opt , i.e. the higher secondary throughput can be achieved by the
~ opt [ C
and C
CCSS
C-CSS scheme compared to the T-CSS scheme.
~ opt increases rst and then decreases with
In the T-CSS scheme, the value of C
the increase of N. In fact, when N increases from 1 to 2, due to the diversity, the
cooperative spectrum detection performance is far greater than the single point
detection performance, and at this time, the false alarm probability of Qopt
will be
f
opt
~
greatly reduced, and so the value of C rapidly increases. When the N continues to
increase, the occupied bandwidth for sending local decision results will increase,
~ opt , and then C
~ opt reduce thereafter.
which have a great influence on the value of C
While in the C-CSS scheme, the value of MC will be relatively increase with the
increase of N, in order to ensure the target detection probability, Popt
d;CCSS must be
opt
small, i.e. we should set a higher detection threshold kN;CCSS to achieve a lower
false alarm probability of Popt
f ;CCSS . However, whether the false alarm probability
Qopt
is
increased
or
not
depending
on the influence degree by MC and Popt
f ;CCSS
f ;CCSS .
opt
~ opt
decreases with the increase of N, while C
As shown in Fig. 4.9b, Q
CCSS
f ;CCSS
basically increases.
Figure 4.10 illustrates the performance comparison between the C-CSS and the
T-CSS scheme under different channel idle probability of P0 by using the optimal
algorithm. The normalized secondary throughput and the false alarm probability
varied with the number of sensing users N are shown in Fig. 4.10a, b respectively. It
can be seen that the false alarm probability of Qopt
and secondary throughput of
f
opt
~
C in the T-CSS scheme are not influenced by the channel idle probability of P0.
While the false alarm probability of Qopt
f ;CCSS in the C-CSS scheme increases with
opt
~ opt
the increase of P0. The secondary throughput of C
CCSS is affected by both Qf ;CCSS
opt
opt
~
and MC. The smaller of MC the larger of C
, and the larger of Q
the
CCSS
f ;CCSS
~ opt
~ opt . The detailed status of C
smaller of C
CCSS
CCSS is shown in Fig. 4.10a, and
opt
Qf ;CCSS is shown in Fig. 4.10b.
Figure 4.11 shows the performance comparison between the C-CSS and the
T-CSS scheme with reporting error (Pe = 0.1) and without reporting error (Pe = 0).
The normalized secondary throughput and the false alarm probability varied with
the number of sensing users N are shown in Fig. 4.11a, b respectively. It can be
seen that, when there is no reporting error, the false alarm probability of C-CSS is
greater than that of T-CSS scheme. When N > 2, due to the fact that MC \M, thus
~ opt . When the reporting error exists, the false alarm probability of
~ opt [ C
C
CCSS
C-CSS is greatly reduced and the secondary throughput is greatly improved compared to the T-CSS scheme.
92
4.5
Summary
In this chapter, we analyzed the C-CSS performance under both imperfect sensing
channels and reporting channels. An optimal detection algorithm has been proposed
for maximizing the access opportunities of SUs from the point of secondary user.
The analytical results show that the control bandwidth can be effectively saved, the
dependence of reporting error to sensing user number can also be reduced by
C-CSS. The tail sensing problem can be effectively solved and the false alarm
probability also decreased. The maximized spectrum access chance of secondary
user can be realized while assuring target detection probability by optimal algorithm. In addition, C-CSS performance from the point of secondary user throughput
is analyzed for further. The relationship among SU, detection threshold, bandwidth,
spectrum utilization and secondary throughput is studied and the optimal algorithm
for maximizing the secondary throughput is proposed. The research results indicate
that the required detection threshold for maximizing the secondary throughput is
different between traditional CSS and C-CSS under given SU number while
assuring object detection probability. The larger secondary throughput can be
achieved with less SU in C-CSS. The optimal secondary throughput can be realized
by the proposed optimal algorithm.
References
cognitive radios. In: Proceedings 38th Asilomar conference signals, systems, computers.
2. Sahai A, Hoven N, Tandra R (2004) Some fundamental limits on cognitive radio. In:
Proceedings of Allerton conference signals, systems, and computers. Monticello
3. Helstronm CW (1988) Improved multilevel quantization for detection of narrowband signals.
IEEE Trans Aerosp Electron Syst 24(2):142147
4. Hashlamoun WA, Varshney PK (1996) Near-optimum quantization for signal detection. IEEE
Trans Commun 44:294297
5. Blum RS (1999) Distributed detection for diversity reception of fading signals in noise. IEEE
Trans Inf Theory 45:58164
6. Chamberland JF, Veeravalli VV (2003) Decentralized detection in sensor networks. IEEE
Trans Sig Process 51:407416
under bandwidth constraints. In: IEEE wireless communications networking conference
8. Rago C, Willett P, Bar-Shalom Y (1996) Censoring sensors: a low communication-rate
scheme for distributed detection. IEEE Trans Aerosp Electron Syst 32(2):554568
9. Jiang RX, Chen B (2005) Fusion of censored decisions in wireless sensor networks. IEEE
Trans Wireless Commun 4(6):26682673
10. Yiu S (2009) Nonorthogonal transmission and noncoherent fusion of censored decisions. IEEE
Trans Veh Technol 58(1):263273
Proc IEEE 97(5):878893
References
93
in cognitive radio networks-[transaction letters]. IEEE Trans Wireless Commun 7(12):4761
4766
13. Li ML, Yuan CW, Li L et al (2011) Performance analysis and optimization of cooperative
spectrum sensing for maximizing secondary throughput. Tongxin Xuebao 32(2):5360
14. Li ML, Yuan CW, Li L et al (2012) Performance analysis and optimization of censoring based
cooperative spectrum sensing. J Xidian Univ 39(1):115121
Chapter 5
CSS Technology with Relay
In this chapter, another CSS technology is shown, which utilize relay to improve
the CSS performance. The system model and some optimal algorithm of CSS are
stated in the following. The simulation results are shown in each section.
5.1
Introduction
Cooperative transmissionwhere the original idea comes from the basic relay
model that consists of three terminals: a source, a relay, and a destinationis well
known as a powerful technology that combats signal fading due to multipath
propagation in a wireless medium. By enabling a set of cooperating relays to
forward received information, this regime exploits spatial diversity through cooperation among distributed antennas belonging to multiple terminals in wireless
networks [1, 2]. Therefore, cooperative relay technology has attracted widely
attention for increasing transmission diversity gain. Cooperative diversity [3, 4] can
effectively combat the channel fading and enhance the throughput, which however
at the expense of a reduction in spectrum efciency. In order to improve the
spectrum utilization, the relay selection based cooperative diversity has been
investigated in [57], where only the best relay is selected to forward a source
nodes signal. All the research papers mentioned above address the conventional
non-cognitive radio network. As stated in [8], the relay technique has been generally considered as an effective method to improve the capacity and coverage for
next generation wireless networks, which is also effective in cognitive radio networks. In [8], the authors have considered a secondary transmitter to act as a relay
for primary transmissions. It has been shown that the secondary link throughput can
be improved in certain network topologies. The CSS performance based on relay
cooperation is investigated in [9, 10], which focus on how to cooperate between
SUs. Paper [1] has investigated the use of cooperative relay to assist the fulllment
of heterogeneous trafc demands in a secondary network with an unbalanced
DOI 10.1007/978-3-319-31095-4_5
95
96
5 CSS Technology with Relay
spectrum usage. Paper [1] has proposed a scheme to improve diversity by cooperative relay in CRN. More recently, papers [1116] investigate the relay spectrum
sensing protocol including multiple relay and best relay, which mainly focus on the
optimization of sensing overheads and the enhancement of the received signals by
relays to improve the sensing performance.
In traditional (non-cognitive radio) multiple-relay networks, three relay protocols
(i.e., xed relaying, selection relaying and incremental relaying) have been studied
extensively in [2, 17]. The advantages of such relaying protocols are achieved at the
cost of a reduction in spectral efciency, since the relays are used to transmit on
orthogonal channels to avoid interfering each other. While in actual CRN, there
exist multiple relays which can assist SUs transmission. If every available relay
participates in the cooperation, the system resource will be costed largely, meanwhile the system interference will also be increased such as the interference
between SUs and relays, as well as between relays and PUs. To address the
shortcomings of inefcient utilization of the spectrum resource, a best-relay
selection protocol has been investigated in [5, 6, 1820], where only the best
relay is selected to forward a source nodes signal and thus only two channels (i.e.,
the best relay link and direct link) are required regardless of the number of relays. It
has been shown in [5] that the best-relay selection scheme can achieve the same
diversity-multiplexing tradeoff performance as the traditional protocols where all
relays are involved in forwarding the source nodes signal. Accordingly, the
best-relay selection is also an attractive relay protocol for cognitive radio networks
due to its spectrum efciency.
In the research about the cooperative spectrum sensing with best relay above, the
relay, which can successfully decode the transmitted signal and can achieve the
highest received SINR at the destination, is viewed as the best one. Therefore
from the point of improving the transmission reliability in the reporting channels, it
is not adequately considered about the channel condition from SU to a relay in the
existing schemes. In CSS, the signals received by the FC will be handled fairly. In
fact, some of the SUs local detection results may be incorrect due to their imperfect
sensing channels, which may affect the nal decision made at the FC and degrade
the overall sensing performance. In another hand, when SUs reporting channels are
in heavy shadowing the ultimate sensing results made at the FC will be also
degraded due to the reporting errors. It is obvious that when both of the sensing
channel and the reporting channel of one SU are deeply faded, its nearly no
signicance for the SU to take part into cooperation. For the two previous situations, we should make full utilization as long as one of the channels is better. As a
result, we utilize relay cooperation to alleviate the reporting error and improve the
overall sensing performance and propose an object based best relay cooperation
spectrum sensing scheme (Pe-BRCS), in which the best relay is selected by minimizing the probability of reporting error to improve the sensing performance. The
simulation results show that the proposed scheme can make the reduced reporting
error probability and the improved sensing performance compared with the cooperative spectrum sensing scheme with best relays by maximizing the received
SINR.
5.1 Introduction
97
In addition, in most of the existing cooperative spectrum sensing method they

consider the condition of full cooperation, i.e. whether the target SU requires the
relays cooperation, the relay will assist. The unnecessary signaling overhead
undoubtedly increases when the SU does not need the relays assist. In order to save
the system expenses and avoid the unnecessary resource waste, we propose an
adaptive cooperative spectrum sensing scheme with best relay (ABRCS), where,
the SU will adaptively decide whether the relays cooperation needed according to
its reporting channel condition. The simulation results show that the same sensing
performance could be achieved by ABRCS as Pe-BRCS whether SUs reporting
channel is good or not.
5.2
5.2.1
System Model
Problem Description
As shown in Fig. 2.10 in Chap. 2, during each cooperative spectrum sensing

process, there are two essential phases: (1) sensing phase, where all SUs attempt to
detect the presence of a PU individually; and (2) reporting phase, where each SU
sends its initial detection result to the FC such that the FC can make a nal decision
on the presence of the PU by using a given fusion rule, such as AND, OR and so on
[2125]. To assure sufcient protection to the PU, we utilize OFDMA model for
transmission that is, each SU sends its initial detection result to FC on orthogonal
frequency bands to solve the interference problem induced by longer sensing
duration. Cooperative spectrum sensing has been shown to greatly improve the
sensing performance with sensing user increased from 1 to 2. With the increasing of
sensing diversity order, more control bandwidth will be occupied for reporting the
local sensing results which could make the spectrum utilization reduced. Moreover,
the sensing performance could not be obviously improved when sensing diversity
order increases to a degree [24, 25]. Consequently, it is not worth to achieve little
performance gain at the expense of much more system resource. In another hand, as
shown in Chap. 3, when the sensing channel is deeply faded, the sensing performance could not be improved by increasing sensing users, and the lower bound of
false alarm probability increased multiply with the sensing diversity order. In fact,
in the cooperative spectrum sensing process described above, not all of the SUs
local detection results are benecial to FCs nal decision. For example, some of
the SUs local detection results may be invalid to FCs nal decision due to their
imperfect sensing channels, in which it is not necessary for SU to send the local
detection results so as to save signaling costs. On the other hand, it is also
insignicant for SUs transmission to FC when their reporting channels are in heavy
shadowing. Whereas, some useful sensing information will be lost if we discard
those SUs whos one of the sensing channel and reporting channel condition are
favorable. To fully utilize SUs available information, we propose a cooperation
spectrum sensing scheme with relay.
98
Fig. 5.1 CSS model without

relay
SU1
PU
FC
SU2
Fig. 5.2 CSS model with

relay
SU1 (target SU)

PU
FC
SU2 (SR)
For instance, as shown in Fig. 5.1, in traditional CSS, SU1 and SU2 implement
local spectrum sensing independently and then transmit the results to the FC.
The FC makes a nal decision on the presence of the PU by using a given fusion
rule, such as AND, OR and so on. If the sensing channel of SU2 is deeply faded, the
local decision result may not correctly reflect the actual status of the PU. The
transmitted data from SU2 is unhelpful to the FCs nal decision. However, if the
reporting channel of SU2 is good and the reporting channel of SU1 is severely
faded, we can utilize SU2 as a relay to transmit the local detection result of SU1 to
the FC to improve the transmission reliability as shown in Fig. 5.2. For convenience, we remark the SU who only play as a relay to assist other SUs transmission
as secondary relay (SR). Recently, the research on relay based CSS (R-CSS) get
widely attention. However, the actual users demand was not considered in the
proposed scheme in [1116], in which, no matter how the reporting channel condition of SU1 is, SU2 will forward the signal from SU1 to the FC. In fact, if the
reporting channel condition of SU1 is rather better, it is unnecessary to request SU2
for cooperative spectrum sensing to save signaling cost. In addition, in [1116], not
all of the impact factors have been considered when selecting the best relay, which
is very important to the system performance.
5.2 System Model
99
Fig. 5.3 Process of relay

based cooperative spectrum
sensing
In actual CRN, there exists multiple relays which can assist SUs transmission. If
every available relays participate in the cooperation, the system resource will be
costed largely, meanwhile the system interference will also be increased such as the
interference between SUs and relays, as well as between relays and PUs. It has been
shown in [5] that the best-relay selection scheme can achieve the same
diversity-multiplexing tradeoff performance as the traditional protocols where all
relays are involved in forwarding the source nodes signal. Accordingly, we consider cooperative spectrum sensing with best-relay.
5.2.2
Signal Model
The process of relay based cooperative spectrum sensing (R-CSS) considered in

this paper is shown in Fig. 5.3. Each transmission link between any two nodes as
shown in Fig. 5.3 is modeled as Rayleigh fading and, moreover, the fading is
viewed as constant during one whole time slot. The additive white Gaussian noise
(AWGN) at all receivers has the same power spectral density N0. where, CSF
denotes the average signal to noise ratio of the channel between the SU and the FC,
CiSR is the average signal to noise ratio of the channel between the SU and ith SR,
CiRF is the average signal to noise ratio of the channel between ith SR and the FC
(i = 1, 2, , M; M is the number of SR for assisting SUs transmission). Besides, let
PPU and PSU denote the transmit powers of PU and SU, respectively. For notational
convenience, let HPU denote whether PU is active or not, namely HPU = H0 represents the absence of the PU with probability P0 and HPU = H1 presents its
presence with probability P1. Throughout this paper, we assume that the primary
user status (i.e., presence or absence) does not change during one time slot. Note
that this assumption is applicable to most of the existing medium access protocols,
even for a random access protocol. This is because that many random access
protocols are based on a time slot structure (e.g., slotted ALOHA, slotted CSMA,
and so on), which are more efcient than the corresponding non-slotted protocols.
For such slotted random access protocols, the primary user is present in a
slot-by-slot manner. The R-CSS process includes four steps.
100
Step 1: SU and ith SR receive the signal transmitted by the PU, and the signal
received at SU and ith SR can be expressed as
(
p
PPU hPS xP nPS 1
p i
i
ySR 1 PPU hPR xP niPR 1
ySU 1
5:1
where the index 1 represents the rst sensing phase of current time slot.
Moreover, hPS and hiPR are the fading coefcient of the channel from the
PU to the SU and the ith SR, nPS 1 and niPR 1 are both additive white
Gaussian noise with zero mean and variance N0, and xP is dened as

xP
0; HPU H0
sp ; HPU H1
5:2
where, sp is the transmit signal of the PU in the rst phase of current time
slot. Based on the received signal as given in Eq. (5.1), SU and ith SR
decide whether the PU is active or not, and the corresponding decision is
i
^ SR
^ SU 1 and H
1.
referred to as an initial detection result denoted by H
Step 2: The SU broadcasts a signal xSU , ith SR and the FC receive the signal.
Here, we consider the relay cooperation only. Then, the corresponding
received signal at ith SR and FC can be written as
(
p i
PSU hSR xSU niSR 2
p
ySF 2 PSU hSF xSU niSF 2
yiSR 2
5:3
where, the index 2 represents the second sensing phase of current time
slot. Moreover, hiSR and hSF are the fading coefcient of the channel from
the SU to SRi and the FC, niSR 2 and niSF 2 are both additive white
Gaussian noise with zero mean and variance N0, and xSU is dened as

xSU
^ SU 1 H0
0; H
^ SU 1 H1
xS ; H
5:4
where xS is the encoded signal by the SU after receiving the signal of the
PU.
From Eq. (5.3), ith SR and the FC attempt to decode the signal xSU and
perform CRC checking. If the channel capacity is below a required data
rate, an outage event is said to occur and the decoder fails to recover the
original signal no matter what decoding algorithm is adopted. In this
paper, we dont consider this case and only take into account the occasion
decoded successfully.
5.2 System Model
101
Step 3: The signal received at the FC is expressed as

yiRF 3
q
PiSR hiRF xiSR niRF 3
5:5
where, the index 3 represents the third sensing phase of current time slot.
PiSR denote the transmit powers of ith SR. hiRF is the fading coefcient of
the channel from ith SR to the FC. niRF 3 is additive white Gaussian noise
with zero mean and variance N0, and xiSR is dened as
xiSR
xSU ; decoded successfully

0;
decoded failure
5:6
Step 4: The FC combines the received signal from the SU and ith SR using the
maximum ratio combining (MRC) method, and gives an estimation of the
original signal. The reason for choosing MRC is that it can achieve a
better performance compared with the other two combining methods, i.e.,
the selective diversity combining (SDC) and the equal gain combining
(EGC). Then, from Eqs. (5.3) and (5.5), the combined SINR at the FC is
given by
2 2
PSU jhSF j2 jxSU j2 PiSR hiRF xiSR
SINRF
N0

CSF CiRF ; decoded successfully
CSF ;
decoded failure
5.3
5.3.1
5:7
Proposed Best Relay CSS Scheme

Problem Description
Throughout this paper, we utilize energy detector to evaluate the spectrum sensing
performance, and assume both the sensing channels and reporting channels as
Rayleigh fading channels. Then Over Rayleigh fading channels, the average probability of detection, the average probability of miss detection and the average
probability of false alarm are denoted as Pd , Pm and Pf respectively, which are
shown in Chap. 2. We also suppose that the probability of reporting error is identical
for all SUs as Pe . Then, the average probability of miss detection Pm and the average
probability of false alarm Pf after reporting channels could be calculated as
102
P m 1 P d 1 Pd 1 Pe 1 Pd Pe
P f Pf 1 Pe 1 Pf Pe
5:8
At the FC, all 1 bit decisions are fused together according to the or rule and a
nal decision is made. The false alarm probability of CSS and the miss detection
probability of the CSS are then given by
(

N

N
Qf 1 1 P f 1 1 Pe 1 2Pe Pf
Qm Pm 1 Pe Pe 1 Pm N

N1
@Qf
N 1 Pe 1 2Pe Pf
1 2Pe
@Pf

N1
N 1 Pe 1 2Pe Pf
1 2Pe
5:9
5:10
Here, we assume that the reporting error is less than 0.5 in actual occasions.
@Q
Therefore, we get @Pff 0.
So Qf is monotonically increased with the local false alarm probability Pf . So,
there exists minimal value of Qf
Qf ;minjPf 0 1 1 Pe N NPe
5:11
In CSS, the system detection performance will be degraded by imperfect

reporting channels, i.e. false alarm bound exists in detection performance curves.
Equivalently, the probability of miss detection will drastically increase to one at the
false alarm bound, when cooperative spectrum sensing will be impractical. From
Eq. (5.11), we know that the false alarm bound is influenced by the two factors:
sensing diversity order of cooperative spectrum sensing N and the probability of
reporting error Pe . It is known that the larger sensing diversity order makes the
improved sensing performance, which will however induce the increased lower
bound and the larger false alarm probability. On the other hand, when sensing
diversity order is given, the larger the probability of reporting errors is, the larger
the bound of false alarm probability. Consequently, the performance of cooperative
spectrum sensing is limited by the probability of reporting errors. To deal with this
limitation, we propose a relay cooperation transmission to reduce the reporting
errors. In actual CRN, there exists multiple relays which can assist target SUs
transmission. If every available relays participate in the cooperation, the system
resource will be costed largely, meanwhile the system interference will also be
increased such as the interference between SUs and relays, as well as between
relays and PUs. It has been shown in [18] that the best-relay selection scheme can
achieve the same diversity-multiplexing tradeoff performance as the traditional
protocols where all relays are involved in forwarding the source nodes signal.
Then, how to select the best relay refers to whether the optimal system performance
can be achieved.
5.3 Proposed Best Relay CSS Scheme
5.3.2
103
The SINR-BRCS Scheme
In general, the relay, which can successfully decode the transmitted signal and can
achieve the highest received SINR at destination, is viewed as the best one. As a
consequence, the best relay cooperation scheme by maximizing the received SINR
(SINR-BRCS) can be written as
Best relaySINR arg max
i2Rm

SINRF arg max CSF CiRF
i2Rm
5:12
As we all know, the average signal to noise ratio CSF between the SU and the FC
is invariant when the SU is given. From Eq. (5.12), the best relay selection relies on
the selected relays reporting channel condition CiRF . Obviously, we can see that it
is not enough for considering the channel condition between ith SR and the FC
only. In fact, the channel condition between the SU and ith SR also affects the
cooperative performance. Therefore, not all of the factors which will make influence to the cooperative performance are considered in the SINR-BRCS scheme.
The reason is that the memory-less channel is considered in the scheme, in which
only the source and the destination nodes are assumed and the intermediate procedure is ignored. As a matter of fact, the more accurate detection results can be
achieved only if each step is fully considered in CSS.
5.3.3
Proposed Pe_BRCS Scheme
According to the drawback described above, we propose an object based cooperative spectrum sensing scheme with best relay (Pe-BRCS), in which the best relay is
selected by minimizing the probability of reporting error to improve the sensing
performance.
As described in Sect. 5.2, as a selected SR, it will receive the transmitted signal
from the target SU at the second step of the R-CSS. When the received signal
strength of the SR larger than a threshold kSR , it will perform decode and forward
operation and therefore achieve cooperation transmission for the target SU, otherwise, the target SU will directly transmit its local results to the FC. As a result, the
total reporting error probability of the R-CSS includes two parts: error probability
for direct transmission and error probability for cooperation transmission. Let Pe ,
Pe SF and Pe coop denote the three kind of error probability described above, then,
the following expression can be obtained:
Pe PCSF c0 Pe
SF
PCSF \c0 Pe
coop
5:13
104
Under Rayleigh fading channel [26], the probability density function of signal to
noise ratio can be expressed as pk; C C1 ek=C , where, C is the signal to noise
ratio of the channel from source to destination. Hence, the probability of decode and
forward for SR can be calculated as
Z1
PCSF c0
Pk; CSF dk e
C 0
SF
5:14
c0
When BPSK modulation is applied, the error probability from the target SU to
the FC and SR can be calculated as
Z1
Pe
SF
0
r

p
1
1
CSF
erfc cPc; CSF dc
1
2
2
CSF 1
Z1
Pe
SR
r

p
1
1
CSR
erfc cPc; CSR
1
2
2
CSR 1
5:15
It is seen that, after SR performs decode operation to the received signal from the
target SU, the decoding results may be correct or not. When the incorrect decoding
result is re-encoded and transmitted to the FC by the SR, we assume that the error
probability of decoding by the FC is less than 1/2. When the correct decoding result
is re-encoded and transmitted to the FC, the FC will perform combining and
decoding operation. Under Rayleigh fading channel, probability density function of
c=CFC
signal to noise ratio for MRC is P
c c e
, C is the average signal to
MRC
C2FC
FC
noise ratio. Therefore, when MRC is used, the error probability of decoding by the
FC can be calculated as:
Z1
Pe
MRC
r

CFC 32
p
1
1
CFC
erfc cPMRC cdc
1

2
2
CFC 1
CFC 1
5:16
From Eq. (5.12), when the SR signal is decoded correctly by the FC, we have
CFC CSF CRF :
Consequently, the error probability for cooperation transmission can be calculated as:
Pe
coop
1
Pe
2
SR
1 Pe
SR
Pe
MRC
5:17
105
Combining Eqs. (5.13)(5.17), the total reporting error probability of the R-CSS
can be calculated as
r
r

c
1 c0
CSF
1
CSR
0
1 e CSF 1
Pe e CSF 1
2
8
CSF 1
CSR 1
r
r

c
CSR CRF 32
1
1
CSR
CSR CRF
0
1 e CSF 1

1
2
2 CSR 1
CSR CRF 1
CSR CRF 1
5:18
Let Pie denote the total reporting error probability of the R-CSS when SRi is
selected, then the total reporting error probability of the R-CSS by the proposed
Pe-BRCS scheme can be expressed as
Pe
BRCS
minPie
5:19
Consequently, the best relay by the proposed Pe-BRCS scheme can be written
as:
Best relay arg minPie
i2Rm
5.3.4
5:20
Complexity Analysis
According to the analysis in Sect. 5.3.2, it is needed to calculate every reporting

error probability corresponding to all the relays so as to select the best relay in the
Pe_BRCS scheme. Therefore, certain computational complexity should be
increased in the Pe_BRCS scheme compared to the SINR-BRCS scheme. From
Eq. (5.18), we can see that there exists explicit expressions of Pie , which does not
need any iterative operation. From the simulation results later, we can see that,
compared to the SINR-BRCS scheme, the better performance can be achieved by
the Pe_BRCS scheme under the conditions of a little increase in computation
complexity.
5.3.5
Detection Performance
Considering in Fig. 5.3, the SU and ith SR using energy detection method to
perform local spectrum sensing, i.e. the SU and ith SR will use the detected PU
signal energy value as the judgment standard to make initial judgment. We assume
that the sensing channels of the SU and ith SR are independent and identically
distributed, the calculating method of the initial local detection probability of Pd
106
and the false alarm probability of Pf are described in Sect. 3.3. When the reporting
error probability is denoted as Pe, after the reporting channels transmission, the
calculation expressions of the detection probability P d of the SU and the false alarm
probability P f of ith SR can be presented as follows
P d Pd 1 Pe 1 Pd Pe

P f Pf 1 Pe 1 Pf Pe
5:21
The FC uses the or fusion rule to merge the two received signals and to make a
nal judgment, the specic way is shown in Sect. 3.3.
According to the above analysis, the main difference between the proposed
scheme and the traditional scheme lies in that the different reporting error probability by different optimal relay selecting rules. Ultimately, how to select the relay is
very important. In the following, we will give the process of the relay selecting
algorithm in this paper.
(1) Initialization: when calculating the reporting error probability Pe, the decision
thresholds and the decode-and-forward thresholds should be initialized at each
SU and ith SR, including kiSR , SF and R. In addition to the above parameters,
the time bandwidth product of u and the average signal noise ratio of SNR in
the sensing channels also need to be initialized when calculating the detection
probability and the false alarm probability. Because when the target SU is
determined, the SUs reporting channel condition is certain, it is also needed to
initialize the average signal noise ratio of CSF in the reporting channel.
(2) Generate SRs with different channel conditions of CiSR and CiRF randomly.
(3) According to the previous analysis, calculate the reporting error probability.
For different schemes, the selected relay is different, which in calculation are
mainly embodied in the difference of CiSR and CiRF . In the SINR-BRCS
scheme, the selected relay is the SR which can be used to maximize the value
of CiRF in step (2). In the Pe-BRCS scheme, the relay which can minimize the
reporting error probabilities corresponding to all of the relays is nally being
selected. In realization, the Newton-Raphson method is used to nd out the
relay for minimizing the reporting error probability, which can be referenced
as in Sect. 3.4.
(4) According to the reporting error probabilities obtained by different schemes in
the step (3), the detection probability and the false alarm can be obtained by
Eq. (5.21).
5.3.6
Simulation Results
Figure 5.4 shows the performance comparison between the proposed object based
cooperative spectrum sensing scheme with best relay (Pe-BRCS) and the
107
10
the reporting error probability
NO-RELAY
ARBITRARY-RELAY
SINR-BRCS
Pe-BRCS
-1
10
-2
10
-10
-5
10
15
the threshold of relay decode and forward
Fig. 5.4 The reporting error probability versus the threshold of relay decode and forward
SINR-BRCS scheme. For illustrating the advantages of the proposed scheme

preferably, we also give the scheme without relay (NO-RELAY) and with arbitrary
relay(ARBITRARY-RELAY). In the simulation, we consider the decode and forward threshold to the received signal of each node as 0 dB, CSF 5 dB, CiSR and
CiRF are created randomly, varied from 0 to 20 dB. From the Fig. 5.4, we can see
that the reduced reporting error probability can be achieved by the proposed PeBRCS scheme, however, only little performance improvement by the SINR-BRCS
and ARBITRARY-RELAY scheme compared with NO-RELAY scheme.
Figure 5.5 shows the detection performance comparison among different
schemes. In the simulation, we consider the average SNR of sensing channels is
10 dB, the decode and forward threshold to the received signal of each node as
0 dB, the time and bandwidth product is 5.
It is seen that the miss detection probability and the lower false alarm bound can
be obviously reduced by the proposed Pe-BRCS scheme compare to other schemes
including the SINR-BRCS scheme, the ARBITRARY-RELAY scheme and
NO-RELAY scheme. Besides, we also give the traditional CSS with 2 users
without relay for comparison. It is seen that compared with the traditional CSS with
2 users without relay, the proposed Pe-BRCS scheme can make the reduced lower
false alarm bound which makes SU more chance to access.
108
miss detection probability
100
10-1
NO-RELAY
ARBITRARY-RELAY
SINR-BRCS
Pe-BRCS
CSS with 2 user
10-2
10 -2
10 -1
10 0
false alarm probability
Fig. 5.5 The miss detection probability versus the false alarm probability
5.4
5.4.1
Proposed C-BR-CSS Scheme

Problem Description
In Sect. 5.3, we have proposed an object-based cooperative spectrum sensing with

best relay to improve the system performance. Although the system performance
can be effectively improved by relay cooperation, the system cost and complexity
can also be increased compared without relay. Therefore, we should consider the
actual demand to determine whether the relays cooperation is needed. In order to
avoid the unnecessary resource waste and save the unnecessary signaling cost, we
further propose a censoring-based CSS scheme with best relay(C-BR-CSS), in
which, only when the SU (substitute SU1 in Fig. 5.2) has detected the PU signal and
its reporting channel is deeply faded, the target SU will transmit its local detection
result to the selected best secondary relay (SR) and request for cooperating transmission. It is needed to note that in this kind of centralized-relay-based CSS, the FC
makes a nal decision on the presence of the PU and then broadcast it in the whole
network. Then, in the proposed scheme, the comprehensive information of SU1 and
SU2 was utilized to improve the system performance, i.e. SU1 with a relatively
better sensing channel and worse reporting channel and SU2 with a relatively better
reporting channel and worse sensing channel cooperate with each other to implement spectrum sensing, SU2 can correctly get the nal results decided by the FC.
5.4 Proposed C-BR-CSS Scheme
109
Fig. 5.6 Time slot structure

of C-BR-CSS scheme
5.4.2
System Model
Figure 5.6 depicts a slotted structure of C-BR-CSS. The total sensing time is
composed by three parts: local sensing time ss , which is the same for the SU and
SR; the reporting time-slot of the SU (sSU ) and the reporting time-slot of the SR
(sSR ). In order to assure the detection reliability to the PU, let
sSU maxsSUFC ; sSUSR , where, sSUFC and sSUSR are the occupied time for
reporting local result from the SU to the FC and SR respectively.
We also assume that a SU (SR) has the knowledge of the instantaneous channel
gain of the link from itself to the FC or SR (when sending terminal is a SU) as in
Ref. [27]. In this kind of censoring scheme, when the PU signal is not detected, the
target SU will not report any result to the FC neither to the SR, so the signaling cost
will be greatly saved. It is worth noting that we take more attention to the sensing
phase other than data transmission phase, so, it is very important for how to reduce
the total sensing time as soon as possible. Although the proposed scheme is not
optimal for all occasions, it makes operation simplied, and doesnt need additional
resource for feedback to look for the optimal solution. When the interested spectrum is unoccupied in a relatively long time, it can be continuously used by the SU
until the PU is detected, while no need to vacant the spectrum at a xed time as in
timely spectrum sensing to implement spectrum handoff so as to reduce the handoff
probability and signaling cost.
The C-BR-CSS process includes three steps. Each channel between any two
nodes as shown in Fig. 5.2 subjects to Rayleigh fading and, moreover, the fading is
viewed as constant during one whole time slot. The additive white Gaussian noise
(AWGN) at all receivers has the same power spectral density N0. Besides, let PPU
and PSU denote the transmit powers of the PU and SU respectively. For notational
convenience, let HPU denote whether the PU is active or not, namely HPU = H0
represents the absence of the PU with probability P0 and HPU = H1 presents its
presence with probability P1. Throughout this paper, we assume that the primary
user status (i.e., presence or absence) does not change during one time slot [16].
Then, in the rst step of the C-BR-CSS, the received signal of the SU and SR can
be expressed as
(
p
PPU hPUSU xPU 1 nSU
p
ySR 1 PPU hPUSR xPU 1 nSR
ySU 1
5:22
where the index 1 represents the rst sensing phase of the current time slot, hPUSU
and hPUSR are the channel fading coefcient from the PU to the SU and SR, nSU
110
and nSR are both additive white Gaussian noise with zero mean and variance N0, and
xPU is dened as

xPU
0;
sp ;
HPU H0
;
HPU H1
where, sp is the transmit signal of the PU in the rst phase of current time slot.
Based on the received signal as given in (5.22), the SU and the SR decide whether
the PU is active or not, and recode the signal. The corresponding decision is
^ SR 1 respectively
^ SU 1 and H
referred to as an initial detection result denoted by H
in the rst sensing step. The recoded signal can be denoted as

xSU 1

xSR 1
where,
0;
sSU 1;
^ SU 1 H0
H
^ SU 1 H1
H
5:23
0;
sSR 1;
^ SR 1 H0
H
^ SR 1 H1
H
5:24
H0 ; represents the absence of the PU by the SU

,
H1 ; represents the presence of the PU by the SU

^ SR 1 H0 ; represents the absence of the PU by the SR :
H
H1 ; represents the presence of the PU by the SR
In this sub-section, we consider not only the two-way relay in traditional networks,
but also pay attention to the links between the primary networks and the secondary
networks, which is very important in CRN. Consequently, in the second step of the
C-BR-CSS, the target SU will send the re-encoded signal to the SR and the FC,
meanwhile, the SR and the FC will also receive the interference signal from the PU.
Then, the corresponding received signal at the SR and the FC can be written as
^ SU 1
H
p
p
PPU hPUFC xPU 2 PSU hSUFC xSU 1 nFC
p
p
ySR 2 PPU hPUSR xPU 2 PSU hSUSR xSU 1 nSR
yFC 2
5:25
where, the index 2 represents the second sensing phase of the current time slot.
hPUFC is the channel fading coefcients from the PU to the FC, hSUSR and hSUFC
are the channel fading coefcients from the SU to the SR and the FC, nFC is the
additive white Gaussian noise with zero mean and variance N0, and xSU 1 is
dened as

xSU
^ SU 1 H0
H
^ SU 1 H1 ;
sSU ; H
0;
and sSU is the transmit signal of SU.
111
From (5.23), the SR and FC attempt to decode the signal xSU and perform Cyclic
Redundancy Check (CRC) checking. If the channel capacity is below a required
data rate, an outage event is said to occur and the decoder fails to recover the
original signal no matter what decoding algorithm is adopted. In this paper, we
dont consider this case and only take into account the occasion that the received
signal in each node is decoded successfully. The re-encoded signal at the SR can be
expressed as

xSR 2
sSU 1;
sSR 2;
^ SR 1 H0
H
^ SR 1 H1
H
5:26
where, sSR 2 is an enhanced signal compared to sSU 1, which referred as the CC

mode in the relay cooperative communication.
In the third step of C-BR-CSS, the signal received at the FC is expressed as
yFC 3
p
p
PPU hPUFC xPU 3 PSR hSRFC xSR 2 nFC
5:27
where the index 3 represents the third sensing phase of current time slot. PSR is the
transmit power of the SR. hSRFC
is the channel fading coefcient from the SR to
xSU ; decoded successfully
:
the FC. xSR is dened as xSR
0;
decoded failure
5.4.3
5.4.3.1
Detection Performance Analysis
As shown in Fig. 2, the SU and SR perform local spectrum measurements independently and then make a binary decision respectively, decide whether the PU is
active or not. Here, we adopt the assumption in [25], compared with the distance
from any SU(SR) to the PU, the distance between any SU and SR is small, so that
the received signal at the SU(SR) experiences almost identical pathloss. Therefore,
it is reasonable to assume that the primary signals received at the SU and SR are
independent and identically distributed (i.i.d.) with the instantaneous SNRs in every
link between the PU and SU (or SR when SU acted as a relay) the same mean c.
Furthermore, we assume that all SUs use the same detection threshold . Over
Rayleigh fading channels, the average false alarm probability Pf and the average
detection probability Pd of the SU and the SR are given by Letaief and Zhang [25].
112
Cu; k=2
Cu
u2
X
1
k=2n 1 c=cu1
pd ek=2
n!
n0
"
#
u2
X
1
n
kc=21 c
ek=21 c ek=2
n!
n0
pf
5:28
where u is the time-bandwidth

product, Ca; x is the incomplete gamma function
R1
given by Ca; x x ta1 et dt, and Ca is the gamma function.
After initial detection, only when the target SU has detected the PU signal, it will
send its local result to the FC, and when necessary it will send it to the SR for
cooperation at the same time. In the second sensing step, the SR will decode the
^ SR 1. Here, we take
signal transmitted from the target SU, and combine with H
decision or fusion as example, and assume the decode decision result of the SR to
^ SRSU , then the ultimate decision result made by the SR can be
the target SU as H
^ SR 1 H
^ SRSU , which will be sent to the FC. It is valuably
^ SR 2 H
denoted as H
noted that, in this kind of censoring sending scheme, the signaling cost will be
greatly reduced, especially in the lower density network occasion of the PU. In the
end, the FC combines the received signal from the SU and the SR using or fusion
criterion as example in this paper, and the ultimate decision result can be expressed
^ FC H
^ SU 1 H
^ SR 2.
as H
SU
SR
Let pd;pro and pd;pro denote the local detection probability of the target SU and
SR
the SR, pSU
f ;pro and pf ;pro denote the local false alarm probability of the target SU and
the SR, which can be obtained by (5.28). Pe SF and Pe RF denote the reporting
error probability from the target SU to the FC and the SR to the FC. P SU
d;pro and
SR
P d;pro denote the detection probability of the target SU and the SR after their
SR
imperfect reporting channel respectively, P SU
f ;pro and P f ;pro denote the false alarm
probability of the target SU and the SR after their imperfect reporting channel
SR
SU
respectively. Then, in the proposed C-BR-CSS scheme, P SU
d;pro ,P d;pro , P f ;pro and
P SR
f ;pro can be calculated as:
SU
P SU
d;pro pd;pro 1 Pe
SF
1 pSU
d;pro Pe
SF
5:29
SR
P SR
d;pro pd;pro 1 Pe
RF
1 pSR
d;pro Pe
RF
5:30
SU
P SU
f ;pro pf ;pro 1 Pe
SF
1 pSU
f ;pro Pe
SF
5:31

SR
P SR
f ;pro pf ;pro 1 Pe
113
RF
1 pSR
f ;pro Pe
RF
5:32
So under imperfect reporting channel, the total detection probability Qd;pro and
the false alarm probability Qf ;pro in the proposed C-BR-CSS scheme can be calculated as follows:

Qd;pro 1 1 P SU
1 P SR
d;pro
d;pro
5:33
SR
Qf ;pro 1 1 P SU
1

P
f ;pro
f ;pro
5:34
SU
From the above analysis, we obtain, pSU
d;pro pd , pf ;pro pf . Since we only
consider the occasion that the data transmitted from the target SU to the SR was
decoded successfully, that is, the received signal strength of the SR larger than a
threshold kSR . Under Rayleigh fading channel, the probability density function of
signal to noise ratio between the target SU and the SR can be expressed as
pk; CSR C1SR ekSR =CSR . So the successful decoding probability Pdec of SR can be
R1
SR
calculated as Pdec k pk; CSR dk ekSR =CSR , then, pSR
d;pro and pf ;pro can be
SR
calculated as:
pSR
d;pro 1 1 pd 1 pd Pdec
5:35

pSR
1 pf Pdec
f ;pro 1 1 pf
5:36
While in traditional R-CSS (T-R-CSS) scheme, the target SU will transmit a

enhanced signal to FC by combining the local detection result of itself and the
decoded signal transmitted from the SR, which can be expressed as
^ SU 1 H
^ SUSR . According to the characteristic of channel symmetric and the
H
independence and identically distributed (i.i.d.) of the SU and the SR, we assume
that the decode decision threshold of the SU and the SR is same. The detection
SU
probability pSU
d;tra and the false alarm probability pf ;tra of the SU in T-R-CSS scheme
can be calculated as follows:
pSU
d;tra 1 1 pd 1 pd Pdec
5:37

pSU
1 pf Pdec
f ;tra 1 1 pf
5:38
After the imperfect reporting channel, the total detection probability Qd;tra and
the false alarm probability Qf ;tra in T-R-CSS scheme can be calculated as follows:
114
Qd;tra pSU
d;tra 1 Pe
SF
Qf ;tra pSU
f ;tra 1 Pe
SF
1 pSU
d;tra Pe
SF
5:39
1 pSU
f ;tra Pe
SF
5:40
From (5.30), (5.35) and (5.37), the following function can be obtained
SU
P SR
d;pro pd;tra 1 Pe
RF
1 pSU
d;tra Pe
RF
5:41
According to the denition of fusion criterion, the following expression can be

obtained from (5.33).
Qd;pro P SR
d;pro
5:42
In our proposed scheme [28, 2931], consider the relay with better reporting
channel as the SR, and in another words, just because of the imperfect reporting
channel, the target SU needs relays cooperation, then we get that Pe SF Pe RF .
The purpose of doing research on CSS detection performance is to improve the
spectrum utilization by SUs access opportunistically while assuring sufcient
protection to the PU, then, we consider the actual local detection probability no less
than 1/2, according to (5.39), (5.41) and (5.42), we get
Qd;pro pSU
d;tra 1 Pe
SF
1 pSU
d;tra Pe
SF
Qd;tra
5:43
From the above analysis, it can be concluded that the higher detection probability can be achieved by the proposed C-BR-CSS scheme.
5.4.3.2
Secondary Throughput Analysis
For a given frequency band of interest, let us dene P1 as the probability of the PU
being active, and P0 as the probability of the PU inactive. Then P0 + P1 = 1. There
are two scenarios for which the secondary network can operate at the PUs frequency band.
Scenario I: When the PU is not present and no false alarm is generated by the
SU, the probability for which happen is P0 1 Qf .
Scenario II: When the PU is active but it is not detected by the SU, and the
probability for which happen is P1 1 Qd .
Denote d0 as the throughput of the secondary network when it operates in the
absence of PUs, and d1 as the throughput when it operates in the presence of PUs.
Then, the achievable throughput of the secondary link can be calculated as
115

T ss sSU sSR
T ss sSU sSR
d0 P0 1 Qf
d1 P1 1 Qd
T
T
5:44
In actual occasion of spectrum sharing, what we care is how to increase the

spectrum utilization opportunity for the SU under the PU is protected sufciently,
so as to improve the secondary throughput. According to Ref. [32], for a given
d , the rst item in the right hand side of (5.44)
target probability of detection, Q
dominates the achievable throughput. Therefore, for convenience, the normalized
achievable throughput in secondary link can be written as

~ T ss sSU sSR 1 Qf k
C
T
5:45
where, k denote the local detection threshold, Qf is the corresponding false alarm
probability.
From (5.43), for a given detection threshold k0 , we have Qd;pro k0 Qd;tra k0 .
In the traditional CSS scheme without relay, the detection probability Qd and the
false alarm probability Qf are monotonically decreased with the detection threshold.
From the above analysis, as we all know the influence of detection threshold to the
detection performance in R-CSS is the same as in the traditional CSS scheme
without relay. Consequently, we get
@Qd
0
@k
@Qf
0
@k
5:46
Let kpro denote the demand detection threshold to satisfy the target detection
d in the proposed C-BR-CSS scheme, and ktra is in the T-R-CSS
probability Q
scheme. From (5.43) and (5.46), the expression of kpro ktra can be obtained.
d , the following
In the same theory, to satisfy the target detection probability Q
expression can be obtained
Qf ;pro Qf ;tra
~ pro C
~ tra
C
5:47
Therefore, the lower false alarm probability can be obtained by the proposed
C-BR-CSS scheme under same protection to the PU as the scheme in [16], which
makes the increased access chance for the SU and improved achievable secondary
throughput.
116
-0.1
Detection probability
10
-0.2
10
-0.3
10
C-R-CSS,Pe-SF =0.001
C-R-CSS,Pe-SF =0.01
-0.4
10
C-R-CSS,Pe-SF =0.1
T-R-CSS,Pe-SF =0.001
-0.5
10
T-R-CSS,Pe-SF =0.01
T-R-CSS,Pe-SF =0.1
-3
10
-2
10
-1
10
10
False alarm probability
Fig. 5.7 Detection performance comparison between C-R-CSS and T-R-CSS
5.4.4
Simulation Results
In this section, the simulation results include two parts. Figures 5.7 and 5.8 consist
of the rst part, in which, the advantages of the proposed C-BR-CSS scheme were
illustrated compared to the traditional R-CSS scheme (T-R-CSS). The second part
includes Fig. 5.9, in which, the advantages of the proposed censoring based CSS
scheme with best relay were given.
Figure 5.7 shows the detection performance comparison between the proposed
C-R-CSS scheme and the T-R-CSS scheme under different reporting error probability of SU (Pe SF ), in which sensing parameters are given as kSR 0 dB,
CSR 10 dB, CRF 10 dB, the average signal and noise ratio of sensing channels
SNR = 10 dB, time and bandwidth product u = 5. It is seen that, for a given false
alarm probability, the detection probability can be effectively improved by the
C-R-CSS scheme compared with T-R-CSS scheme. The performance advantages of
C-R-CSS scheme becomes obvious with the increase of reporting error probability
of the target SU, this is because when the reporting error probability of the target
SU (Pe SF ) is larger, the target SU will request for SRs assist retransmission to
achieve the diversity gain in C-R-CSS scheme, while in T-R-CSS scheme, although
the data sending from the SU is enhanced, the ultimate local detection results are
still transmitted by the channel from the SU to the FC which is deeply faded.
Figure 5.8 illustrates the secondary throughput performance comparison
between the proposed C-R-CSS scheme and T-R-CSS scheme under different
sensing channel condition, while assuring the target detection probability
117
0.5
0.45
C-R-CSS,SNR=0dB
C-R-CSS,SNR=5dB
C-R-CSS,SNR=10dB
T-R-CSS,SNR=0dB
T-R-CSS,SNR=5dB
T-R-CSS,SNR=10dB
Secondary throughput
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0
10
12
14
16
18
The reporting channel condition from SU to FC
Fig. 5.8 The secondary throughput performance comparison between C-R-CSS and T-R-CSS
0.5
0.45
NO-RELAY
ARBITRY-RELAY
SINR-BRCS
Proposed
Secondary throughput
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
10
12
14
16
18
The reporting channel condition from SU to FC
Fig. 5.9 Secondary throughput performance comparison under different schemes
d 0:99, in which the simulation parameters are set as ss 0:1T, sSR 0:1T. It
Q
is seen that, the improved secondary throughput can be achieved by the C-R-CSS
scheme compared with the T-R-CSS scheme, in addition, the larger secondary
118
throughput can still be obtained by the former when the target SUs reporting
channel condition is bad such as CSF 0 dB, which is also due to the relays
retransmission.
Figure 5.9 illustrates the secondary throughput versus the reporting channel
condition from the target SU to the FC (CSF ), assuring the target detection prob d 0:99. It is seen that the larger secondary throughput can always be
ability Q
achieved by the proposed scheme. And from the above three gures, we can see
that whether the relay based CSS performance is good or not mainly depends on the
relays selection.
5.5
5.5.1
Proposed Adaptive CSS Scheme with Best Relay

Problem Description
In the previous section, we discuss the cooperation spectrum sensing with best
relay. In fact, as shown in Fig. 5.3, when CSF is higher, in other words, when the
reporting channel condition between the SU and the FC is preferable, the SU can
make the relative accurate detection by itself without the help of relay. If the relay
cooperation is used at this time the channel resource will be wasted undoubtedly. In
order to save the system expenses and avoid the unnecessary resource waste, we
propose an adaptive cooperative spectrum sensing scheme with best relay
(ABRCS). When CSF is higher, the SU transmits its local detection result to FC
without SRs help, in which the signaling cost can be saved between the SU and ith
SR, also ith SR and the FC. Besides, the interference of SRi to PU and SU can also
be reduced. However, when CSF is lower, that is, when the reporting channel
condition between the SU and the FC is rather poor, the SU requests for a SRs
help. As an illustration of the proposed scheme, we further give the specic
implementation as follows. Because of the broadcasting characteristic of wireless
channel, ith SR and the FC will both receive the signals from the SU. Then, it is
hard for ith SR to decide when to decode the received signal by itself. As a
transmitted protocol, a indicate tag can be added in the transmitted data from the SU
besides the actual signal value. For example, when the reporting channel condition
between the SU and the FC is poor, let tag = 1 represents the need for relays
help. Then, the best relay can be selected by the proposed scheme as described in
Sect. 5.3.2. When the reporting channel condition between the SU and the FC is
better, let tag = 0 represents no need for relays help. Then, those SR with better
reporting channels but weak sensing channels can maintain silent at current time
slot to save signaling cost. However, how to evaluate CSF to determine whether the
relay cooperation is needed should be studied for further. If the threshold of CSF is
selected too big, the unnecessary signaling expenses will be wasted, whereas, if it is
too small, the inaccurate sensing results would happen which makes reduced
sensing performance. Consequently, the threshold of CSF will influence the
5.5 Proposed Adaptive CSS Scheme with Best Relay
119
performance of the proposed algorithm directly. The exact threshold of CSF , tagged
as CSF threshold in the following, can be obtained by solving the equation
Pe SF Pe BRCS , that is
Pe
SF

min Pidec
SR

Pie
dec

e 1 Pie

dec
Pie

MRC

1 Pidec

SR
Pe
SF
5:48
5.5.2
Algorithm Design
In order to illustrate the parameters affecting the root of Eq. (5.48), we give the
following equation

gCSF min f1 CiSR f2 CiSR e 1 f2 CiSR f3 CiSF ; CiRF

1 f1 CiSR gCSF
5:49

where, gCSF Pe SF , f1 CiSR Pidec SR , f2 CiSR Pie dec , f3 CiSF ; CiRF
Pie MRC :
We can see that, it is hard to solve the Eq. (5.49) directly. Because what we
discover for CSF threshold is based on the minimal error probability Pe BRCS , which
is relative to relays selection. In other words, only the best relay is given, the
optimal Pe BRCS can be obtained. So we cant consider CiSR and CiRF as constant
simply. Besides, the reporting error probability bringing in relay is not only related
to CiSR but also CSF . For this reason, when CSF is not assured, the optimal Pe BRCS
cant be obtained rstly then to implement adaptive algorithm. In this paper, the
proposed ABRCS is processed as follows:
(1) Initialization.
(2) Creating ith SR (i = 1, 2, N) with different CiSR and CiRF randomly.
(3) For each SR, compute the corresponding reporting error probability when it is
acted as a relay for SUs transmission.
(4) Find the optimal CSF threshold by Newton-Raphson method. Based on the
analysis above, several existing algorithms (Newton-Raphson, Bi-Section,
SECOND) may be used to nd the root of the equation Pe SF Pe BRCS .
However, the Newton-Raphson method needs the explicit expression of
derivative of function, and it has faster converging rate than the other algorithms. From Eq. (5.20), Pe BRCS cant be exactly expressed by the function of
CSF , so it is hard to utilize Newton-Raphson method directly to nd the
optimal CSF threshold . In order to exploit the Newton-Raphson method to
accurately nd the optimal CSF threshold , we make some process to the algorithm as follows:
120
Firstly, as there exists complementary error function erfc in the expression

of Pe SF and Pe BRCS , which includes the integral variables and multi
parameter variables, the Newton-Raphson method cant be utilized directly.
For this purpose, we complicate the argument of hyperbolic tangent function
and make the complementary error function approximate with hyperbolic
tangent function quanticationally [33]. Then, the simplied approximate
result is erfc x 1 tanh1:12838x 0:10277x3 .
Secondly, because CiSR and CiRF are invariant with the relay having been
selected, we should get the optimal reporting error probability for each SR
rstly, and then utilize the Newton-Raphson method for optimal CSF threshold .
Let
gCSF Pe SF CSF Pe BRCS CSF .
The
process
of
the
Newton-Raphson algorithm is listed as follows
Step 1: choose tolerance and the initial guess CSF 1, let i = 1;
Step 2: if jgCSF ij\e, stop; otherwise, go to step 3
Step 3: let CSF i 1 CSF i gCSF i=g0 CSF i. Then, let i = i + 1
and go to step 2.
(5) As to an arbitrary SU in CRN, decide whether the formula is true, that is
CSF CSF threshold ? If it is true, let tag = 0, representing no need for relays
help; otherwise tag = 1 representing the needness for relays help and the relay
based CSS can be completed as mentioned in Sect. 4.3.3.
5.5.3
Simulation Results
Figure 5.10 illustrates the reporting error probability utilized by proposed ABRCS
scheme. To illustrate the advantages of the proposed scheme preferably, we also
give the scheme without relay (NO-RELAY). In the simulation, we consider the
decode and forward threshold to the received signal of each node as 0 dB,
the average SNR of sensing channels is 10 dB. The average SNR from the SU to the
SR and from the SR to the FC are randomly created varied from 20 to 20 dB. It is
seen that no matter how the channel condition of CSF , the minimized reporting error
probability can always be realized by proposed ABRCS scheme.
In Fig. 5.11, we illustrate the sensing performance comparison among
NO-RELAY, the Pe-BRCS scheme and the ABRCS scheme under different average
signal and noise ratio of sensing channels SNRS. It is seen that the optimal sensing
performance can be achieved by ABRCS, same as Pe-BRCS scheme.
5.5 Proposed Adaptive CSS Scheme with Best Relay
121
the reporting error probability
10
ABRCS
NO-RELAY
Pe-BRCS
-1
10
-2
10
-3
10
-20
-15
-10
-5
10
15
20
signal and noise ratio between SU and FC
Fig. 5.10 The reporting error probability versus signal and noise ratio between SU and FC
10
10
-1
NO-RELAY,SNRS =20dB
Pe-BRCS,SNRS =20dB
ABRCS,SNRS =20dB
NO-RELAY,SNRS =10dB
Pe-BRCS,SNRS =10dB
ABRCS,SNRS =10dB
NO-RELAY,SNRS =0dB
Pe-BRCS,SNRS =0dB
10
ABRCS,SNRS =0dB
-2
-2
10
-1
10
Fig. 5.11 The miss detection probability versus the false alarm probability
10
122
5.6
Summary
In this paper, we propose an object based cooperative spectrum sensing scheme

with best relay (Pe-BRCS) under both Rayleigh channel model of sensing channels
and reporting channels. In order to avoid the unnecessary resource waste and save
the unnecessary signaling cost, we further propose a censoring-based CSS scheme
with best relay(C-BR-CSS), in which, only when the SU has detected the PU signal
and its reporting channel is deeply faded, the target SU will transmit its local
detection result to the selected best secondary relay (SR) and request for cooperating transmission. Specially, we propose an adaptive cooperative spectrum sensing
scheme with best relay (ABRCS) further. The simulation results demonstrate that
the proposed Pe-BRCS scheme could make the reduced reporting error probability
effectively and the improved sensing performance compared with SINR-BRCS
scheme. It is also shown that the same sensing performance could be achieved by
ABRCS as Pe-BRCS, while more signaling costs saved by the former. It is noted
that, for analysis conveniently, although the scenario considered is with the single
cognitive source-destination pair, it can be applied and extended to a more generic
scenario with multiple cognitive source-destination pairs. To be specic, if a random access strategy is considered, we can allow multiple cognitive sources to
independently sense the licensed channel with the assistance of multiple relays, the
analysis of which i somewhat complex and is our next work plan. Secondly, when
each node decodes the received signal and makes CRC checking, outage event
would be happen, the analysis of which deserves for further research.
References
1. Zhang Q, Jia JC, Zhang J (2009) Cooperative relay to improve diversity in cognitive radio
networks. IEEE Commun Mag 47(2):111117
2. Nosratinia A, Hunter TE, Hedayat A (2004) Cooperative communication in wireless networks.
3. Laneman JN, Tse DNC, Wornell GW (2004) Cooperative diversity in wireless networks:
efcient protocols and outage behavior. IEEE Trans Inf Theory 50(12):30623080
4. Hunter TE, Sanayei S, Nosratinia A (2006) Outage analysis of coded cooperation. IEEE Trans
Inf Theory 52(2):375391
5. Bletsas A, Shin HM, Win Z, et al (2006) A simple cooperative diversity method based on
network path selection. IEEE J Sel Areas Commun 24(3):659672
6. Beres E, Adve RS (2008) Selection cooperation in multi-source cooperative networks. IEEE
Trans Wireless Commun 7(1):118127
7. Ikki S, Ahmed MH (2010) Performance analysis of adaptive decode-and-forward cooperative
diversity networks with best-relay selection. IEEE Trans Commun 58(1):6872
8. Simeone O, Bar-Ness Y, Spagnolini U (2007) Stable throughput of cognitive radios with and
without relaying capability. IEEE Trans Commun 55(12):23512360
References
123
10. Ganesan G, Li Y (2007) cooperative spectrum sensing in cognitive radio, part II: multiuser
11. Zou YL, Yao YD, Zheng BY (2010) Outage probability analysis of cognitive transmissions:
the impact of spectrum sensing overhead. IEEE Trans Wireless Commun 9(8):26762688
12. Zou YL, Yao YD, Zheng BY (2011) Cognitive transmissions with multiple relays in cognitive
radio networks. IEEE Trans Wireless Commun 10(2):648659
13. Zou YL, Yao YD, Zheng BY (2011) A cooperative spectrum sensing based cognitive relay
transmission scheme without a dedicated sensing relay channel in cognitive radio networks.
IEEE Trans Signal Process 59(2):854858
14. Zou YL, Yao YD, Zheng BY (2011) A selective-relay based cooperative spectrum sensing
scheme without dedicated reporting channels in cognitive radio networks. IEEE Trans
Wireless Commun 10(4):11881198
15. Zhu J, Zou YL, Zheng BY (2009) Cooperative detection for primary user in cognitive radio
networks. EURASIP J Wireless Commun Networking. doi:10.1155/2009/617320 [Online].
Available: http://downloads.hindawi.com/journals/wcn/2009/617320.pdf
16. Zou YL, Zhu J, Zheng BY et al (2010) An adaptive cooperation diversity scheme with
best-relay selection in cognitive radio networks. IEEE Trans Signal Process 58(10):54385445
17. Hunter TE, Nosratinia A (2006) Diversity through coded cooperation. IEEE Trans Wireless
Commun 5(2):283289
18. Nosratinia A, Hunter TE (2007) Grouping and partner selection in cooperative wireless
networks. IEEE J Sel Areas Commun 25(2):369378
19. Seddik KG, Asdek AK, Su WF et al (2007) Outage analysis and optimal power allocation for
multimode relay networks. IEEE Signal Process Lett 14(6):377380
20. Lkki S, Ahmed MH (2008) Performance of multiple-relay cooperative diversity systems with
best relay selection over Rayleigh fading channels. EURASIP J Adv Signal Process, Article ID
580368
cognitive radio networks. In: IEEE international conference on communications, China,
pp 34113415
(12):47614766
Proc IEEE 97(5):878893
26. Zhao Y, Adve R, Lim TJ (2005) Outage probability at arbitrary SNR with cooperative
diversity. IEEE Commun Lett 9(8):700702
27. Zou YL, Yao YD, Zheng BY (2012) Cooperative relay techniques for cognitive radio systems:
spectrum sensing and secondary user transmissions. IEEE Commun Mag 50(4):98103
28. Li ML (2013) Performance analysis and optimization of cooperative spectrum sensing with
relay. Tongxin Xuebao 34(9):3340
29. Li ML, Wang AH (2013) A object based cooperative spectrum sensing scheme with best relay.
IEICE Trans Fundam/Commun/Electron/Inf Syst E96-A(6):14921495
30. Li ML, Yuan CW, Li L et al (2013) Target based cooperative spectrum sensing scheme with
best relay. J Jilin Univ 43(4):10981103
31. Li ML (2012) A best relay based cooperative spectrum sensing scheme. Appl Mech Mater
157:14921495
32. Liang YC, Zeng YH, Peh ECY et al (2008) Sensing-throughput tradeoff for cognitive radio
33. Rifths DV, Smith IM (1991) Numerical methods for engineers. Blackwell Scientic, London
Chapter 6
CSS Based on Soft Combination
The CSS based on hard combination is analyzed in the previous chapter. In this chapter,
the CSS based on soft combination is illustrated, which can be used to improve the CSS
performance by introducing weighting factors. In the following, the system model and
some optimal algorithms are shown from theory to simulation results respectively.
6.1
Introduction
The detection performance of spectrum sensing schemes is usually limited by

destructive channel conditions between the primary users and the secondary users.
Because of the channels deep fading, it is very difcult to distinguish between a white
spectrum and a weak signal. The authors have proposed cooperative transmission
among secondary users exploiting spatial diversity in order to improve the reliability
of spectrum sensing [1, 2]. In such scenarios, the secondary users in CRN experience
different channel conditions from the primary users, if they combine the sensing
information jointly, it will have a better chance of detecting the primary users. In other
words, cooperative spectrum sensing can improve the detection performance by
exploiting spatial diversity, and thus reduce the probability of interfering with primary
users. Since cooperative spectrum sensing is generally coordinated over a separate
control channel, efcient cooperation schemes should be designed to reduce bandwidth and power requirements while maximizing the sensing reliability.
The CSS includes two successive processes: sensing process and reporting
process. In the traditional research on the CSS based on energy detection, the
hard-combining method is considered in data fusion, i.e. the energy signal collected
by each SU is transmitted to a fusion centre (FC) via a dedicated control channel.
Recently, the soft combination based CSS brings much attention, in which the
weighting factor is introduced before transmitted data to make full use of the best
information to improve the detection performance [3].
DOI 10.1007/978-3-319-31095-4_6
125
126
6 CSS Based on Soft Combination
However, It has been shown that cooperative spectrum sensing needs a control
channel for each SU to report its sensing result and the control channel is usually
bandwidth limited [1, 4]. In [5], a non-linear optimization enigma is used to formulate
the CSS problem which might be difcult to implement. In [3], an optimal soft combination scheme is proposed, based on some approximation in the target optimal
function. They develop an efcient linear cooperation framework for spectrum sensing,
where the global decision is based on simple energy detection over a linear combination
of the local statistics from individual nodes. In [6, 7], a linear soft combination of raw
measurements from individual cooperative SUs performed at the fusion center is considered. In Ref. [3, 57], each SU transmits the real value of its local sensing observation, the innite bits will be required and this will result in a large communication
bandwidth when there is large number of SU which will cause a great waste of
communication bandwidth. Quantization of local observations has attracted much
attention even though it may bring additional noise and a signal-to-noise ratio
(SNR) loss at the receiver [4, 8]. A lot of work has been done on the quantization for the
signal detection but most of them focused on the optimal design of the quantizer. It has
been shown that identical binary quantization, i.e., one bit quantization, performs
asymptotically optimal as the number of users goes to innity [9]. We call this as the
CSS based on soft decision fusion, which can save the control bandwidth efciently and
improve the spectrum utilization. Besides, in actual occasion, not only the imperfect
sensing channels exist but also the imperfect reporting channels exist [1013]. However,
in [3, 57], they only consider the channel between the PU and the SU is imperfect.
In this paper, we concentrate on the more actual CSS scenario with both imperfect
sensing channels and reporting channels, in which the CSS performance based on soft
combine with energy fusion (SC-EF-CSS) and decision fusion (SC-DF-CSS) are
analyzed respectively. The expressions of the false alarm probability and the detection
probability are given when introducing the weighting factor as reliability. The optimal
algorithm to obtain the weighting vector for maximizing the detection probability is
given based on N-P criterion. In order to reduce the complexity for realizing the optimal
algorithm based on N-P criterion, the optimal algorithm for obtaining weighting vector
based on modied deflection coefcient (MDC) is given further. The results show that,
only a little detection performance loss exists in SC-DF-CSS compared with CSS based
on soft combine with energy fusion, while the system bandwidth can be effectively
saved by the former. In addition, it is also shown that the detection performance can be
effectively improved by the optimal algorithm, and the optimal algorithm based on
MDC can achieve the same performance with the optimal algorithm based on NP
criterion, which however, makes the computational complexity reduced greatly.
6.2
System Model
The process of CSS based on soft combination is shown in Fig. 6.1. The original
status of PU is assumed as H0 or H1, representing the vacant spectrum present or
not. Each SU i perform local spectrum sensing independently using energy
6.2 System Model
127
Fig. 6.1 The process of CSS based on soft combination
detection and the observation energy signal represented as xi. We assume that the
reporting channel gain is hR;i which is Rayleigh fading, nR;i is the additive white
Gaussian noise with zero mean and variance r2R;i .
In the SC-EF-CSS scheme, the observation result xi is preprocessed by
weighting vector and then transmitted to the FC via reporting channel. While in the
SC-DF-CSS scheme, ith SU performs local spectrum measurements independently
and makes a binary decision ui, which is then transmitted to the FC. yi is the local
decision signal after reporting channel, which is rstly weighed by weighting factor
i and then transmitted to the FC. In the end, all of the received local results are
combined at the FC as yc, based on which, the PUs presence or not is decided.
6.3
SC-EF-CSS Performance
In the existing research on the CSS based on soft combining with data fusion, they
only considered the imperfect sensing channels, and not considered the imperfect
reporting channels. For the convenient comparison, we analyze the performance of
SC-EF-CSS and give the optimal algorithm for selecting the weighting vector in
this sub-section.
6.3.1
1. Performance of single sensing

The spectrum sensing can be regarded as a binary hypothesis testing problem:
H0 : PU is absent
H1 : PU is in operation
The key parameters of the spectrum sensing include the probabilities of error
detection and the probabilities of correct detection. We suppose that there are
128
N independent secondary users in cognitive radio system, whose coverage is far less
than the distance between cognitive radio network and the primary network.
When ith SU performs local spectrum sensing independently, ai t is the
observed signal at the ith SU at time t. The local spectrum sensing problem can be
described as the following hypothesis:

ai t
ni t; H0
hi st ni t; H1
6:1
where, ni t is the additive white Gauss noise (AWGN) in the sensing channel of ith
SU, st is the signal from the PU, hi is the channel gain of the sensing channel
between the PU and the ith SU.
When the ith SU makes a determination based on the local observation, the
energy collected in the frequency domain is denoted by xi . The observation
bandwidth is W and the observation time window is T. Then xi obeys the following
distribution [10, 14]:

Yi
v22u ; H0
v22u 2ci ; H1
6:2
where, u = TW is the time bandwidth product, v22u denotes a central chi-square

distribution with degrees of freedom 2u, v22u 2ci represents a noncentral chi-square
distribution with degrees of freedom 2u, ci is the instantaneous SNR of the received
signal at the ith SU. We assume that the average signal to noise ratio (SNR) in each
link between the PU and SUi as ci . Then the mean and variance of signal xi under
hypothesis H0 and H1 can be calculated as in [15]:

E xi jH0 2ur2s;i
E xi jH1 2u ci r2s;i
6:3
Var xi jH0 4ur4s;i

Var xi jH1 4u 2ci r4s;i
6:4
According to the central limit theorem, over Rayleigh fading channels, the
average false alarm probability of Pf ;i and the average detection probability of Pd;i
can be calculated as
"
Pf ;i
ki E x i j H 0
Pxi [ ki jH0 Q p
Var xi jH0
"
Pd;i
ki E x i j H 1
Pxi [ ki jH1 Q p
Var xi jH1
6:5
#
6:6
6.3 SC-EF-CSS Performance
129
where, r2s;i is the Gaussian variance of ith sensing channel. ki is the local detection
threshold of SUi.
1
Q x p
2p
Z1
2
et =2 dt:
2. Performance of SC-EF-CSS under Imperfect Reporting Channels

In data fusion step, the observation result xi is preprocessed by weighting vector
and then transmitted to the FC via reporting channel. Then, the received signal of
FC can be calculated as:
yi hR;i xi nR;i
6:7
Here, we assume that the reporting channel gain hR;i is identical with the signal
of xi , then the mean and variance of signal yi can be calculated as:

Eyi jH0 E hR;i Exi jH0 E nR;i
Eyi jH1 E hR;i Exi jH1 E nR;i
6:8

Var yi jH0 Var hR;i xi jH0 Var nR;i
Var yi jH1 Var hR;i xi jH1 Var nR;i
6:9

where,
h

2 i
2
Var hR;i xi jH0 E hR;i xi jH0
E hR;i xi jH0
h
i
2
E h2R;i E xi jH0 2 E hR;i Exi jH0 2
h
i

2
Var hR;i xi jH1 E h2R;i E xi jH1 2 E hR;i Exi jH1 2
Here, we assume that the reporting channels subject to the Rayleigh fadings, i.e.
hR;i is characterized by Rayleigh fading, whos kth moment can be calculated as [15]:

k=2
k
C 1
E hkR;i 2d2i
2
6:10
where, di is the Rayleigh fading factor of ith reporting channel. The mean and
variance of reporting channel gain hR;i can be calculated as:
r
p
E hR;i
di
2

p
Var hR;i 2 d2i
2
6:11
130
From Eqs. (6.3), (6.4), (6.8), (6.9) and (6.11), the mean and variance of signal yi
under hypothesis H0 and H1 can be calculated as:
p
2pdi ur2s;i
p
Eyi jH1 2pdi u ci r2s;i
q0;i Eyi jH0

q1;i

p
n20;i Var yi jH0 2 2 d2i ur2s;i 8d2i ur4s;i r2R;i
2

p
n21;i Var yi jH1 2 2 d2i u ci r2s;i 8d2i u 2ci r4s;i r2R;i
2
6:12
6:13
In data fusion step, each signal after reporting yi is pre-processed by weighting

vector xi and then transmitted to the FC via reporting channel. Then, the received
signal of the FC yc can be calculated as
yc
N
X
x i yi WT Y
6:14
i1
Here, the combining weight for the signal from a particular user represents its
contribution to the global fusion. For example, if a SU generates a high-SNR signal
that may lead to correct detection on its own, it should be assigned a larger weighting
coefcient. For those SUs experiencing deep fading or shadowing, their weights are
decreased in order to reduce their negative contribution to the decision fusion.
From Eqs. (6.8), (6.9) and (6.14), the statistic mean and variance of yc under the
two hypotheses can be calculated as:

Eyc jH0 WT P0
Eyc jH1 WT P1
6:15
Var yc jH0 WT D0 W
Var yc jH1 WT D1 W
6:16
h
iT

where, W x1 ; x2 ; . . .; xN T ; Dj diagnj , nj n2j;1 ; n2j;2 ; . . .; n2j;N ; Pj qj;1 ;
qj;2 ; . . .; qj;N T ; j 0; 1.
We assume that kc is the global threshold at the FC. Then, the total false alarm
probability and the total detection probability of SC-EF-CSS scheme can be
evaluated as
"
#
kc WT P0
Qf ;E Q p
WT D 0 W
"
#
6:17
kc WT P1
Qd;E Q p
WT D 1 W
6.3.2
131
The Optimal Algorithm Based on N-P Criterion
In CSS, the main metric of sensing performance is either minimization of the false
alarm probability for a target detection probability or maximization of the detection
probability under given target false alarm probability. In this paper, we mainly
f is given, the
consider the latter [16]. When the target probability of false alarm Q
global decision threshold c can be calculated as

p
WT D 0 W WT P 0
kc Q1 Qf ;E
6:18
where, Q1 is the inverse function of Q.

Therefore, the detection probability for satisfying the target false alarm proba f can be calculated as:
bility Q
"
d;E
Q
p
f
WT D0 W WT P0 WT P1
Q1 Q
p
Q
WT D 1 W
6:19
It is clear that the weighting vector W plays an important role in determining the
PDF of the global test statistic yc. So how to select the appropriate W to maximize
d;E becomes our optimization goal. From
the target detection probability of Q
Eq. (6.19), Q is a monotonically decreased function, therefore, according to N-P
criterion, the optimal algorithm model can be written as
8
<
:
max F W min
W
WT D0 W WT P0 WT P1
p
f
Q1 Q
W D1 W
6:20
s:t: kWk22 1
As to the fact that the solution of formulation (6.20) is difcult to get. We can
solve the following Eq. (6.21). In the following, we prove the minimum value exist
in F W
@F W
0
@W
6:21
We assume the actual occasion of opportunistically access, where, the PU is

protected sufciently, i.e. the detection probability is larger than the false alarm
probability. From Eq. (6.12) the following expression can be obtained:
P1 P0 [ 0
6:22
132
According to the Cauchy inequality, the following expression can be further

obtained:
p
f
WT D0 W WT P0 WT P1
Q1 Q
p
F W
WT D 1 W
p

f
WT D0 W kWk22 kP1 P0 k
Q1 Q
p

WT D1 W
6:23
Therefore, the minimum value of F W exists.

Let
p
f
WT D0 W WT P0 WT P1
F1 W Q1 Q
F2 W
p
WT D1 W
6:24
6:25
Then, the solution of Eq. (6.21) can be obtained as follows

f D0 W
@F1 W Q1 Q
p P1 P0
@W
W T D0 W
6:26
@F2 W
D1 W
p
@W
WT D1 W
6:27
@F W @ F1 W=F2 W
@W
@W
@F1 W
@W F2 W
2 W
@F@W
F 1 W
F2 W2
6:28
From (6.21), (6.24) to (6.28), the solution of Eq. (6.21) can be written as:
Wopt;NP

e0
D1
I eNP D1
0 D1 P1 P0
f 0 K
Q1 Q
6:29
where,
eNP
2
p
p
f e
e0 WT P1 P0 Q1 Q
0
T
2
;
e
W
D
W
;
e
WT D1 W; IK is K
0
0
1
f e
Q1 Q
1
dimensional matrix of one.

According to Ref. [6], Q1e0Q D1
does not affect the value of F W. Therefore,
f 0
we can set it as 1. While, eNP is still the function of W. The optimal weighting
vector Wopt;NP can be calculated by multiple iterations utilizing N-P criterion as in
133
Ref. [7]. Under the constraint of (6.20), the normalized optimal weighting vector is
given by:
Wopt;NP

Wopt;NP
Wopt;NP
6.3.3
The Optimal Algorithm Based on MDC
6.3.3.1
Deflection Criterion
6:30
Let x be a real observation vector and S x the output of an arbitrary receiver. The
2
0 S
deflection associated ith S x is dened by DS E1 SVE
where E1 and E0 are
0 S
the mean expectations under H1 and H0, respectively, V0 is the variance under H0.
The deflection is sometimes called the output signal-to-noise ratio. The deflection
has a long history, predating the likelihood ratio (LR) theory and maximization of
deflection was considered under various assumptions in [1719]. It is worth
pointing out that, in many cases, it has played very important affection by the use of
deflection criterion in the LR receiver. Consider the detection of a deterministic
signal in a Gaussian noise. The LR receiver is then linear and often called the
matched lter [20]. The same lter is obtained by maximizing the deflection
D(S) for systems S(x). Similarly, the LR receiver associated to the detection of a
Gaussian signal in Gaussian noise is quadratic with respect to x. Its structure is the
same as that obtained by maximizing the deflection of quadratic systems S(x). This
was obtained in terms of square law detector and linear ltering in [17] and generalized for quadratic lters in [18]. In conclusion, when Gaussian assumption is
introduced, the maximization of the deflection leads to the same structures as LR
theory. This is a good reason to study the advantages and disadvantages of
deflection in detail in more general situations.
6.3.3.2
Optimal Algorithm
In this section, we utilize a heuristic approach to control the combining weights

[21], which optimizes a modied deflection co-efcient (MDC) that characterizes
the probability distribution function (PDF) of the global test statistic at the fusion
center. This approach slightly compromises the detection performance with less
computational complexity and provides near-optimal solutions for general systems.
According to the analysis above, the equation of (6.19) is difcult to solve to
obtain the optimal weighting vector, whereas, the multiple iterations are needed
which make the higher computing complexity. MDC characterizes the
variance-normalized distance between the centers of two conditional PDFs, which
provides a good measure of the detection performance [6]. Therefore, we introduce
134
MDC to measure the effect of the PDF on the detection performance and to nd the
optimal weighting vector. The expression of MDC d 2 W is dened as:
d 2 W
T
2
W P1 P0
E yc jH1 Eyc jH0 2
E yc jH1
WT D1 W
6:31
Therefore, the optimal target turns to how to select the appropriate weighting
vector W to maximize the value of d 2 W. The optimal model can be expressed as
8
2
<
WT P1 P0
max d 2 W WT D W
1
6:32
W
:
s:t: kWk22 1
Thus, the optimal solution can be obtained by solving the equation
The following solution can be get:
@d 2 W
@
@W
@W
(
WT P1 P0
WT D 1 W
@d 2 W
@W
0:
2 )
h
i2
6:33
T
2P1 P0 T WP1 P0 WT D1 W 2 P1 P0 W D1 W
T
2
T
2
W D1 W
W D1 W
Let (6.33) as 0, then, the optimal weighting vector Wopt;MDC for maximizing the
value of d 2 W can be calculated as:
Wopt;MDC
WT D1 W
D1 P1 P0 eD1
1 P1 P0
WT P1 P0 1
6:34
D1 W
where, although e WW
is the function of W, it does not affect the detection
T
P P
1
performance [6]. So, for analysis conveniently, let e 1. According to formula

(6.32), the optimal normalized weighting vector can be obtained as follows.
Wopt;MDC

Wopt;MDC
Wopt;MDC
6.3.4
6:35
Simulation Results
Figure 6.2 shows the CSS performance using optimal algorithm when only one and
both channels characterized by fading, in which the average SNR of sensing
channels was given as 0 dB, time and bandwidth u = 5, the average noise variance
of sensing channels as 0.5. We also consider that the reporting channel is Rayleigh
135
1
0.9
detection probability
0.8
0.7
0.6
0.5
0.4
0.3
S-R-fade N=1
S-fade N=1
R-fade N=1
S-R-fade N=10
S-fade N=10
R-fade N=10
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Fig. 6.2 The detection performance influence by both imperfect sensing channel and reporting
channel
fading and the fading factor as 1, the average noise variance of reporting channels
as 0.5. Qf denotes the false alarm probability and Qd is the detection probability. It
is seen that, both the sensing channels and reporting channels influence the sensing
1
0.9
0.8
0.7
0.6
0.5
0.4
NO-OPT:N=1
OPT-NP:N=1
OPT-MDC:N=1
NO-OPT:N=2
OPT-NP:N=2
OPT-MDC:N=2
NO-OPT:N=10
OPT-NP:N=10
OPT-MDC:N=10
0.3
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Fig. 6.3 The performance comparison by using optimal algorithm and EGC in the SC-EF-CSS
scheme
136
performance, and the latters influence is larger than the former, which in another
words shows that it is meaningful to consider the influence of the reporting channel
to detection performance.
Figure 6.3 shows the performance comparison by using the two optimal algorithms and EGC method in the SC-EF-CSS scheme under both imperfect sensing
channels and reporting channels. OPT-NP denotes the optimal algorithm based on
N-P criterion, OPT-MDC denotes the optimal algorithm based on MDC criterion,
and EGC denotes the EGC being used for fusion. It is seen that, the performance
by OPT-NP is nearly similar to that by OPT-MDC, and the sensing performance
can be improved by using the two optimal algorithms, and the advantages of the
optimal algorithm become distinct with larger sensing users of cooperation.
6.4
SC-DF-CSS Performance
In this sub-section, we will analyze the performance of SC-DF-CSS, and give the
optimal algorithm for selecting the weighting vector. Because the key difference of
SC-DF-CSS with SC-EF-CSS is that, lie in: after receiving the PU signal, SU does
not directly send the detected data to the FC, it will rstly make a local decision of
1 bit which is then sent to the FC. The specic implementation process refers to the
analysis in the second chapter. Therefore, the analysis of SC-DF-CSS is similar to
that of SC-DF-CSS.
6.4.1
It is different from SC-EF-CSS, in SC-DF-CSS, ith SU performs local spectrum

measurements independently and makes a binary decision ui, and then transmitted
to the FC. In detail, the local decision criterion is
if xi [ ki ; let ui 1
if xi \ki ; let ui 0
6:36
The transmitted signal gi after reporting channels could be represented as

gi
hR;i nR;i ;
nR;i ;
ui 1
ui 0
6:37
6.4 SC-DF-CSS Performance
137
Then the mean and variance of signal gi under hypothesis H0 and H1 could be
calculated as
X
h0;i E gi jH0
pui jjH0 Egi jui j
j0;1
r
p
Pf ;i di
2
r
X
p
E gi jH1
pui jjH1 Egi jui j
Pd;i di
2
j0;1

1 Pf ;i E nR;i Pf ;i E hR;i nR;i
h1;i
f20;i Var yi jH0
f21;i
6:38
pui jjH0 Var yi jui j
j0;1

p
2 Pf ;i d2i r2R;i
X 2
Var yi jH1
pui jjH1 Var yi jui j
6:39
j0;1

p
2 Pd;i d2i r2R;i
2
After reporting channel, each signal gi is pre-processed by weighting vector -i
and then transmitted to the FC via reporting channel. Then, the received signal of
the FC gc can be calculated as
gc
N
X
G
- i gi W
6:40
i1
From Eqs. (6.38) to (6.40), the statistic mean and variance of gc under the two
hypotheses can be calculated as:
(
T H0
E gc jH0 W
T H1
E gc jH1 W
6:41
T R0 W

Var gc jH0 W
T

Var gc jH1 W R1 W
6:42

T
-1 ; -2 ; . . .; -N T ;
W
Hj hj;1 ; hj;2 ; . . .; hj;N ; j 0; 1;
h
iT
Rj diagfj ; fj f2j;1 ; f2j;2 ; . . .; f2j;N ; j 0; 1:
where,
We assume that sc is the global threshold at the FC. Then, the total false alarm
probability and the total detection probability of SC-DF-CSS scheme can be
evaluated as
138
2
T
H0 7
6 sc W
Qf ;D Q4 q 5

T R0 W
W
3
2
T
H1 7
6 sc W
Qd;D Q4 q 5

T R1 W
W
6.4.2
6:43
The Optimal Algorithm Based on N-P Criterion
In order to compare with the SC-EF-CSS scheme, we still consider the optimal
target for maximizing the detection probability under given target false alarm
f is given, the global
probability. When the target probability of false alarm Q
decision threshold sc can be calculated as

sc Q1 Qf ;D
q
T R0 W
W
T H0
W
6:44
Therefore, the detection probability for satisfying the target false alarm proba f can be calculated as:
bility Q
d;D
Q
3
2
q

T
T
T
1

W R0 W W H0 W H1 7
6Q Qf
q
Q4
5

T R1 W
W
6:45
to maximize the target detection

Therefore, how to select the appropriate W

probability of Qd;D becomes our optimization goal. From Eq. (6.45), Q is a
monotonically decreased function, therefore, according to N-P criterion, the optimal
algorithm model can be written as
8
<
:
min
max GW

W
f
Q1 Q
p
T
H0 W
T H1
T R0 W
W
W
p
k 1
s:t: kW
2
2
R1 W

W
6:46
As to the fact that the solution of formulation (6.46) is difcult to get. We can
solve the following Eq. (6.47). In the following, we prove the minimum value exist
.
in GW

@GW
0
@W
6:47
139
We assume the actual occasion of opportunistically access, where, the PU is

protected sufciently, i.e. the detection probability is larger than the false alarm
probability. From Eq. (6.38) the following expression can be obtained:
H1 H0 [ 0
6:48
According to the Cauchy inequality, the following expression can be further

obtained:
q
T R0 W
W
T H0 W
T H1
W

q
G W
T R1 W

W
q

kW
k2 kH 1 H 0 k
T R0 W
f
Q1 Q
W
2
q

T

R1 W
W

f
Q Q
1
6:49
exists.
Therefore, the minimum value of GW
The solution of Eq. (6.47) can be calculated as follows:
@G2 W G1 W

=G2 W
@G1 W G2 W
@ G 1 W
@GW
@W
@W
2

@W
@W
G2 W
6:50
where

Q1 Q
f
G1 W
q
T R0 W
W
T H0 W
T H1 ;
W

G 2 W
q

T R1 W
W
We can get that

f R0 W
Q1 Q
@G1 W
q H1 H0

@W
T R0 W

W
6:51

@G2 W
R1 W
q

@W
T R1 W

W
6:52
From (6.47) to (6.52), the solution of Eq. (6.47) can be written as:
opt;NP
W
where,

r0
1
R1
0 IK rNP R0 R1 H1 H0

Qf
Q1
6:53
140
rNP
2
q
q
T H1 H0 Q1 Q
f r
r0 W
T
0

; IK is
T R1 W

2
W
;
r
;
r
W
W
0
0
1
f r
Q1 Q
1
K dimensional matrix of one.

According to the analysis in Sect. 5.3,
r0
1
f R0
Q1 Q
does not affect the value of
The
. Therefore, we can set it as 1. While, rNP is still the function of W.
GW
opt;NP can be calculated by multiple iterations utilizing
optimal weighting vector W
N-P criterion as in Ref. [7]. Under the constraint of (6.46), the normalized optimal
weighting vector is given by:

W

opt;NP
W
opt;NP
Wopt;NP
6.4.3
6:54
The Optimal Algorithm Based on MDC
Similarly, we introduce MDC to measure the effect of the PDF on the detection
performance and to nd the optimal weighting vector. The expression of MDC
is dened as:
d 2 W
h

d 2 W
Egc jH1 E gc jH0
Var gc jH1
T H 1 H 0
W
i2
T R1 W

W
6:55
Therefore, the optimal target turns to how to select the appropriate weighting
. The optimal model can be expressed as
to maximize the value of d 2 W
vector W
8
<
:
W
max d 2 W
T H1 H0 2

W
s:t:
k2
kW
2
T R1 W

W
6:56
Thus, the optimal solution can be obtained by solving the Eq. (6.57).

@d 2 W
0

@W
6:57
opt;MDC for
According to the analysis above, the optimal weighting vector W
2
maximizing the value of d W can be calculated as:
opt;MDC
W
141
T R1 W

W
T H 1 H 0
W
1
R1
1 H1 H0 rR1 H1 H0
6:58
In the same theory, according to the analysis in Sect. 6.3, Let r 1, then, the
opt;MDC for maximizing the value of d 2 W
can be
optimal weighting vector W
calculated as:

W

opt;MDC
W
opt;MDC
Wopt;MDC
6.4.4
6:59
Simulation Results
Figure 6.4 illustrates the performance comparison between the two optimal algorithm and traditional method under both imperfect sensing channels and reporting
channels. In the simulation, the average SNR of sensing channels was given as
0 dB, time and bandwidth u = 5, the average noise variance of sensing channels as
0.5. We also consider that the reporting channel is Rayleigh fading and the fading
factor as 1, the average noise variance of reporting channels as 0.5. The tag
1
0.9
0.8
0.7
0.6
0.5
0.4
NO-OPT:N=1
OPT-NP:N=1
OPT-MDC:N=1
NO-OPT:N=2
OPT-NP:N=2
OPT-MDC:N=2
NO-OPT:N=10
OPT-NP:N=10
OPT-MDC:N=10
0.3
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Fig. 6.4 The performance comparison between optimal algorithm and traditional algorithm
142
OPT-NP denotes the optimal algorithm based on NP and OPT-MDC denotes the
optimal algorithm based on MDC, NO-OPT denotes the traditional suboptimal
algorithm. It is seen that, the sensing performance is nearly the same by using
OPT-NP and OPT-MDC, and the sensing performance can be improved by using
the two optimal algorithms. It also can be seen that the advantages of the optimal
algorithm become distinct with larger sensing users of cooperation.
Figure 6.5 illustrates the performance comparison between SC-EF-CSS and
SC-DF-CSS by using the optimal algorithm based on MDC. It is seen that, only a
little detection performance loss exists in SC-DF-CSS compared with SC-EF-CSS,
especially in actual occasion, the two performance difference smaller under target
false alarm probably less than 0.1. Besides, it also can be seen that, the false alarm
probability of SC-DF-CSS scheme is less than the one of SC-EF-CSS scheme under
the same target detection probability. In CSS, the false alarm probability is usually
used to evaluate the spectrum utilization. The lower the false alarm probability, the
higher the vacant spectrum can be reused and the spectrum efciency can be
improved. Therefore, the higher spectrum efciency can be achieved by the
SC-DF-CSS scheme compared to SC-EF-CSS.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
SC-EF-CSS:N=1
SC-DF-CSS:N=1
SC-EF-CSS:N=2
SC-DF-CSS:N=2
SC-EF-CSS:N=10
SC-DF-CSS:N=10
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Fig. 6.5 The performance comparison between SC-DF-CSS and SC-EF-CSS
6.5 Summary
6.5
143
Summary
In this chapter, we studied the CSS performance based on soft combination. Firstly,
we analysis the performance of CSS based on soft combining with data fusion
under both imperfect sensing channels and reporting channels, give the optimal
algorithm for selecting the weighting vector. According to the deflection that much
control bandwidth will be occupied in CSS based on soft combination with data
fusion, which leads to the reduced spectrum utilization, in addition, the imperfect
reporting channel has not been considered at present. The actual occasion with both
imperfect sensing channel and reporting channel is considered in this paper. A CSS
based on soft combine with decision fusion (SC-DF-CSS) is proposed and the
detection performance is analyzed for further. At last, the optimal algorithm to
obtain the weighting vector is given. The results show that, only a little detection
performance loss exists in SC-DF-CSS compared with SC-EF-CSS. The detection
performance could be effectively improved by the optimal algorithm, and the
optimal algorithm based on modied deflection coefcient can achieve the same
performance with the optimal algorithm based on N-P criticism, which makes the
complexity reduced greatly.
References
cognitive radios. In: Proceedings of 38th Asilomar conference signals, systems and computers.
United States, pp.772776
3. Zhi Q, Cui S, Sayed AH (2008) Optimal linear cooperation for spectrum sensing in cognitive
radio networks. IEEE J Sel Top Sign Proces 2(1):2840
under bandwidth constraints. IEEE wireless communication and networking conference,
5. Ma J, Zhao GD, Li Y (2007) Soft combination and detection for cooperative spectrum sensing
in cognitive radio networks. IEEE Trans Wireless Commun 7(11):45024507
6. Shen B, Kwak KS (2009) Soft combination schemes for cooperative spectrum sensing in
cognitive radio networks. ETRI, 31(3):263273
7. Shen B, Kwak K, Bai ZQ (2009) Optimal linear soft fusion schemes for cooperative spectrum
sensing in cognitive radio networks. In: Proceedings of 2009 IEEE global telecommunications
conference, GLOBECOM 2009, Honolulu, HI, United states, pp 16
8. Sahai A, Hoven N, Tandra R (2004) Some fundamental limits on cognitive radio. In:
Proceedings of Allerton conference signals, systems, and computers, Monticello
9. Chamberland JF, Veeravalli VV (2003) Decentralized detection in sensor networks. IEEE
Trans Signal Proces 51:407416
Proc IEEE 97(5):878893
144
(12):47614766
12. Li ML et al (2011) Performance analysis and optimization of cooperative spectrum sensing for
maximizing secondary throughput. Tongxin Xuebao 32:5360
13. Li ML et al (2011) Analysis of secondary throughput and optimization in cooperative
spectrum sensing. J China Univ Post Telecom 18:3944
15. Proakis John G (2001) Digital communications, 4th edn. McGraw-Hill, New York
16. Li ML et al (2011) Performance analysis and optimization of cooperative spectrum sensing
based on soft combination. In: 2011 7th international conference on wireless communications,
networking and mobile computing (WiCOM)
17. Gardner WA (1976) Structurally constrained receivers for signal detection and estimation.
IEEE Trans Commun COM-24, 578592
18. Picinbono B, Duvaut P (1988) Optimal linear-quadratic systems for detection and estimation.
IEEE Trans Inf Theory 34:304311
19. Picinbono B, Duvaut P (1990) Geometrical properties of optimal Volterra lters for signal
detection. IEEE Pansactions Inf Theory 36:10611068
20. Van Trees HL (1968) Detection, estimation, and modulation theory, Part I. Wiley, New York
21. Li ML, Dong ZS (2014) CSS performance analysis and optimisation based on soft decision
fusion. Int J Wireless Mobile Comput
Chapter 7
The SS Application in ICIC
In this chapter, an inter-cell interference coordination method by spectrum sensing

technology is studied, where, the terminal with cognitive function can automatically
detect the available spectrum when requesting for service, and select the idle
spectrum or the occupied spectrum with the weakest signal strength to realize the
actual weakest interference communication, make the reduced inter-cell interference and improved cell edge users performance. In the following, the interference
analysis and the solving method are illustrated.
7.1
Introduction
In wireless communication, various kinds of interference in the system restrict the

enhancement of spectrum utilization, service rate and the coverage range abilities,
etc. By using the CR technology, many kinds of the interference can be effectively
reduced such as the self-interference, spectrum sharing interference and the interference between users, etc. Moreover, the favourable wireless communication
system with low interference and without interference can be realized [13].
LTE-Advanced is the evolution version of LTE (Long Term Evolution), which
can satisfy the needs of the future wireless communication market with higher
demands and more applications [46]. The mobile communication network is a
real-time changing dynamic network. Along with the future network evolution,
many challenges will be introduced in the related works of LTE-Advanced when
the personal equipment develops, such as the micro base station, the pico base
station, indoor coverage, the relay station and the home base station, etc [7]. In the
future wireless system, the distribution of system loads, frequency allocation, power
allocation and the interference among systems are all changed. The control and the
management abilities to the base station have also been greatly increased. Most of
the operators hope that some self-organization mechanisms can be introduced to
intelligently adjust the network parameters to optimize the allocation of resources
DOI 10.1007/978-3-319-31095-4_7
145
146
7 The SS Application in ICIC
and reduce the operating costs according to the actual operation of the network [4].
The self organizing network (SON) with self conguration and self optimization
function has been used as a specic requirement in the LTE standard [8], which
requires some advanced functions such as self-planning, self-establishment,
self-deployment, self-allocation, self-operation, self-optimization, self-healing and
other features. These advanced features require that the network should have the
sensing ability to the network environment and can make fully use of the sensing
information [9]. Therefore, the characters with more autonomy and intelligence can
be realized by SON.
In order to improve the network performance and meet the needs of LTE, the
more effective interference management techniques must be used to solve the
interference problem in the future system. Orthogonal frequency division multiplexing (OFDM) technique can guarantee the orthogonality between symbols by
using orthogonal sub-carrier [10, 11], which effectively solve the problem of
interference among symbols. Due to the limited spectrum resources, inter-cell
interference is inevitable even through the OFDM technology is used, which will
seriously affect the cell edge users rate. Then, how to maximize the spectrum
utilization and avoid the mutual interference under the existing limited spectrum
resource environment is one of the main problems to be solved. In addition, the
multiple antenna technology can be used to improve the cell-center users service
rate, however, it is difcult to improve the cell-edge users service rate, which will
result the differences larger between the performance of cell-center and cell-edge
[12]. In the LTE-Advanced system, the interference mainly comes from the users
and the base station in adjacent cells. The interference coordination techniques can
be used as an important means to ensure the system work stable [13].
The cell edge interference can be reduced in a certain extent by using soft
frequency reuse (SFR) technology [1434]. However, in the current static interference coordination techniques, they only consider the application scenarios that
the load of each cell is exactly the same, i.e. each cells loads are synchronously
changed. In fact, each cells service distributions are independent, which has also
been considered in the existing semi-static interference coordination method [35
37], for example, authors tried to meet the different service load requirements by
reserving variable resources. However, in this kind of method, the resources are
limited reserved, which will reduce the spectrum utilization. In addition, in the
existing inter cell interference coordination technologies, although the interference
have been reduced by allocating orthogonal resources for the adjacent cell edge
users in some extent, the interference coordination efciency may be seriously
affected, especially when the service load in the cell center area and in the edge area
are both heavy, the performance of cell edge users is still very poor. The frequency
utilization will become low when the service distribution in the cell center area is
relative to that in the edge area.
Cognitive radio technology as a frequency planning techniques can use the
spectrum sensing to intelligently perceive the surrounding environment. It can
realize the optimal allocation of resources and the improved spectrum utilization by
detecting the use of the radio frequency. Then, in this paper, we propose an
7.1 Introduction
147
inter-cell interference coordination method based on spectrum sensing. The terminal with cognitive function can automatically detect the signal strength of the
spectrum when it intends to launch a new service request, which has been used by
the licensed users, and then it can select the idle spectrum or the spectrum with the
weakest signal to communicate, which can realize the real weakest interference
communication, and reduce the inter-cell interference, enhance the cell edge users
performance.
In this chapter, we rstly analyze the interference of the SFR scheme in the LTE
system, and put forward the method of reducing the inter-cell interference by
spectrum sensing technology. The results show that compared to the SFR scheme,
the proposed scheme can greatly reduce the neighboring co-channel interference,
especially the interference in the uplink, and really improve the performance of cell
edge users and achieve the better interference coordination effect.
7.2
7.2.1
Problem Analysis
The SFR Scheme
In cellular networks, the frequency reuse factor K is usually used to describe the
reuse rate of the spectrum resources [38]. The SFR scheme breaks through the
traditional limit of xed frequency reuse factor, the whole area is divided into the
central area and the edge area with different frequency reuse factor, which are
distinguished by using different transmission powers. The higher powers are allocated to the cell edge users and the lower powers are allocated to the cell center
users. The total frequency reuse factor is related to the transmission power ratio of
the cell center users to the cell edge users, which is changed. When it is equal to 0,
the system frequency reuse factor is 3, and each cells available bandwidth is the
total bandwidth of 1/3; when it is 1, the system frequency reuse factor is 1, and each
cell can use the total bandwidth; when it is between 0 and 1, the frequency reuse
factor in the cell center area and the edge area is different, and the bandwidth can be
used by each cell is between 1/3 and 1 of the total bandwidth.
In the system of the traditional xed frequency reuse factor, the spectrum utilization can be maximized with reuse factor of 1, but the users located at the cell
edge may suffer from more serious co-channel interference. While the system with
reuse factor of 3 can solve the problem, in which, the spectrum utilization is
relatively low. Therefore, the SFR scheme comprehensively consider the cell edge
users co-channel interference and the whole systems spectrum utilization by
power control to change the frequency reuse factor K, and let K vary from 1 to 3
[19, 34].
Therefore, in the SFR scheme, the spectrum resource which each cell can use
varied with the change of the transmit power ratio of center users and edge user.
The smaller ratio means the greater system soft frequency reuse factor and the
148
smaller co-channel interference, which results the lower spectrum utilization. in

another hand, the larger ratio means the smaller system soft frequency reuse factor,
which results the higher spectrum utilization, however, the co-channel interference
of the cell edge users will be more serious.
In the following, we will illustrate the spectrum resource allocation mode and the
power allocation strategy of the SFR scheme as shown in Fig. 7.1.
Let S denotes the total sub-band group, which is divided into three sub-band
groups respectively as S1 ; S2 ; S3 . We assume that each sub-band is continuous. As
shown in Fig. 7.1, all the center users use the low transmission power Pinner , and all
the edge users use the high transmission power Pedge . In the cell tagged as 1, the
cell center users can use the sub-band groups S2 and S3 , and the cell edge users can
use the sub-band group S1 . In the cell tagged as 2, the cell center users can use the
sub-band groups S1 and S3 , and the cell edge users can use the sub-band group S2 .
In the cell tagged as 3, the cell center users can use the sub-band groups S1 and S2 ,
and the cell edge users can use the sub-band group S3 . It is seen that, the sum of the
bandwidth which can be used by the three adjacent cell edge users is equal to the
sum of the total bandwidth, and the sum of the bandwidth which can be used by
each cell center users and edge users is also equal to the total bandwidth.
7.2.2
Interference Analysis
According to the analysis in Sect. 7.2.1, in the SFR scheme, the interference of the
cell center users are mainly from the edge users in the adjacent cells, which are
3
1
2
Power
1
2
2
3
Cell-1
S1
3
1
S2
S3
subbands
1
Cell-2
2
3
S2
S3
subbans
3
1
S1
1
2
Cell-3
S1
S2
S3
subbands
Fig. 7.1 The soft frequency reuse scheme
7.2 Problem Analysis
149
allocated by the high power level, and the interference source only has one at most
from the three adjacent areas. The interference of the cell edge users are mainly
from the center users in the adjacent cells, which are allocated by the low power
level, however there may be three interference sources from the three adjacent
areas. In the uplink, the contribution of the cell center users and the cell edge users
to the base station are different.
When the system is fully loaded, all of the spectrum are used by each cell, the
cell edge users performances can not be improved by SFR, especially when the
trafc of edge area is larger. As shown in Fig. 7.2, because the center user in the cell
tagged as 2 and the edge users in the cell tagged as 1 can use the same spectrum
resource, when the edge user UE1,A and the center user UE2,A located in the
position shown in Fig. 7.2 are both employing the sub-band S11 to communicate. In
the uplink, the service base station of BS1 receives not only the useful signal from
the user UE1,A, but also receives the interference signal from UE2,A. In fact, as
shown in Fig. 7.3, although in the downlink, the two user UE2,A and UE2,B in the
cell 2 will make the same interference to the user UE1,A, in the uplink, obviously,
UE2,B will make the larger interference to the user UE1,A than that of UE2,A, this is
because of the imbalanced resources allocation in the inner-cell, which has not been
considered in the SFR scheme.
d1
Cell 1
UE 1, A
UE 2,A
Cell 2
d2
UE 2,B
Cell 3
Fig. 7.2 The co-channel interference of cell edge users
Cell 1
UE 2,A
UE 1,A
Cell 2
UE 2,B
Cell 3
Fig. 7.3 The co-channel interference of cell center users
150
In the asymmetric services, as shown in Figs. 7.2 and 7.3, UE1,A can use the
same spectrum as that of UE2,B to communicate in the uplink, because of the
relative long distance between them, the uplink interference can be reduced and the
spectrum utilization can be enhanced.
7.3
The ICI Coordination Based on SS
7.3.1
Scheme Design
7.3.1.1
Spectrum Allocation
In this paper, we mainly consider the inter-cell interference in LTE system. The
system resources are divided into N parts with regard to the network structure using
N cell model. In the following, we will take the network structure with 3 cell model
shown in Fig. 7.4 as an example to make a detailed explanation on the initial
allocation rules of the spectrum resources.
The initial spectrum resource allocation mode in the cell is shown in Fig. 7.5.
Firstly, all the frequency
into three sub-band groups S1, S2 and
resources are divided

S3. The resources of Si Si;1 ; Si;2 ; . . .; Si;Ni ; i 1; 2; 3 can be used by the center users
in the ith cell, where the cell i refers to all the cells tagged as i in network
topologies, rather than a certain cell, and the cell edge user can use the resources
which have been allocated to the cell centers users in the adjacent cells with
opposite priority sequences.
Take the network structure shown in Fig. 7.4 as an example, the center users in
cell 1 have priority to use the resources of sub-band group S1, the center users in
cell 2 have priority to use the resources of sub-band group S2, and the center users
in cell 3 have priority to use the resources of sub-band group S3. Similarly, when
allocating the resources for edge users, the edge users in cell 1 have priority to use
the resources of sub-band group S3, the edge users in cell 2 have priority to use the
Fig. 7.4 The network
structure with 3 cells
1
2
3
151
Power
Cell 1
Cell center
S1,1 , S1,2 ,
Cell 3
Cell 2
, S1, N
S 2,1 , S 2,2 ,
, S 2, N
S3,1 , S3,2 ,
, S3, N
Sub-band
Power
Cell edge
S3, N , S3, N 1 ,
3
, S3,1 S1, N , S1, N 1 ,

1
, S1,1 S 2, N , S 2, N 1 ,
2
, S 2,1
Sub-band
Fig. 7.5 The initial frequency allocation rule of each base station
resources of sub-band group S1, and the edge users in cell 3 have priority to use the
resources of sub-band group S2.
In the process of allocating resources for the center users and the edge users,
when resources need to be repetitively allocated, they should be orthogonally
allocated. The cell edge user can use the resources having been allocated to the
center users in adjacent cells in opposite order. For example, we can congure
resources in the priority order of S1;1 ; S1;2 ; . . .; S1;N1 , when allocating the resources
of the sub-band group S1 for the center users in cell 1; and in the priority order of
S1;N1 ; S1;N1 1 ; . . .; S1;1 , when allocating the resources of the sub-band group S1 for
the edge users in cell 2. Similarly, the edge users
in cell 1 can use resources with the

priority order of S3 S3;N3 ; S3;N3 1 ; . . .; S3;1 , the edge users in cell 3 can use

resources with the priority order of S2 S2;N2 ; S2;N2 1 ; . . .; S2;1 .
In the process of initial resource allocation described above, the length N1, N2
and N3 of three sub-bands S1, S2 and S3 are not necessarily equal, which can be set
according to the actual service demand. For example, if the center service in cell 1
is larger than other cells, we can allocate more resources for the center users in cell
1 to meet the actual needs.
When an UE moves to the cell and is ready to launch a new service, the base
station rstly determines whether the UE belongs to the center users or the edge
users, which can be determined by calculating the average carrier-to-interference
ratio or the distance between the user and the base station [36]. When the UE
belongs to the center user, the base station will allocate the resources according to
the second preset frequency allocation rules. For detail, if the service loads are
lower and the cell available initial resources exist, the base station will allocate the
resources for the UE according to the initial frequency resource allocation rules.
Otherwise, the base station will allocate the resources, which can be used by the
adjacent cell center users, for the UE. In the conguration, in addition to guarantee
the orthogonality of the allocated resources with the resources used by the edge
users within the same cell, also with the resources used by the edge users in the
adjacent cell. Take the network structure shown in Fig. 7.4 as example, the second
152
preset frequency allocation rules can be described as Fig. 7.6. When allocating
resources to the center users of cell 1, if the sub-band group S1;1 ; S1;2 ; . . .; S1;N1 has
been completely allocated, the sub-band group of S2 should be prior allocated with
the prior order of S2;1 ; S2;2 ; . . .; S2;N2 ; When allocating resources to the center users
of cell 2, if the sub-band group S2;1 ; S2;2 ; . . .; S2;N2 has been completely allocated,
the sub-band group of S3 should be prior allocated with the prior order of
S3;1 ; S3;2 ; . . .; S3;N3 ; When allocating resources to the center users of cell 3, if the
sub-band group S3;1 ; S3;2 ; . . .; S3;N3 has been completely allocated, the sub-band
group of S1 should be prior allocated with the prior order of S1;1 ; S1;2 ; . . .; S1;N1 .
When the UE belongs to edge users of a cell, it will implement adaptive spectrum sensing and choose the available spectrum with the weakest signal to realize
communication. When the ratio of the bandwidth having been used by the cell
center users to the total bandwidth is less than a threshold, the UE will detect the
available spectrum for communication in the residual bandwidth; otherwise, the UE
will detect the signal strength of the residual bandwidth and select the spectrum
with the weakest signal to realize communication.
The proposed inter-cell interference coordination method based on spectrum
sensing technology can be implemented as the following steps shown in Fig. 7.7.
Through the spectrum detection and selection rules described above, the cell
edge users interference can be effectively solved. The worst inter-cell interference
between adjacent cell edge users can be avoided, and the interference among
adjacent cells can always be weakest. As shown in Fig. 7.3, UE2,A and UE2,B will
both be considered as the cell edge users of cell 2, which have the same distance to
the base station. In the existing inter-cell interference coordination scheme, the edge
users of cell 2 can occupy the same sub-band as the center users of cell 1, in which,
UE2,A will create more interference to the users of cell 1 compared to UE2,B in the
uplink. The weakest interference between UE2,A and the adjacent cell users can be
realized by the proposed scheme.
7.3.1.2
Power Allocation
The hierarchical power resource allocation method is used in the scheme, i.e. the
center users and the edge users are allocated different powers to mitigate inter-cell
interference.
Power
Cell 1
S1,1 , S1,2 ,
, S1, N ; S 2,1 , S 2,2 ,

1
Cell 3
Cell 2
, S 2, N
S 2,1 , S 2,2 ,
, S 2, N ; S3,1 , S3,2 ,
2
, S3, N S3,1 , S3,2 ,

3
, S3, N ; S1,1 , S1,2 ,

3
, S1, N
subband
Fig. 7.6 The second preset frequency allocation rule for cell center users
153
Set the threshold for dividing the cell edge

users and the cell center users
S1
S2
Allocate the frequency resources for users according

the pre-defined frequency resources allocation criterion
Distinguish the new UEs type according the

threshold described above
Decide whether the initiate frequency

resources have been fully allocate
S3
Yes
No
S401
S402
Allocate the frequency resources for the new

UEs according to the second pre-defined
frequency resources allocation criterion
The new UEs described above select the

available spectrum to communicate by
spectrum sensing technology
Allocate power for each UE
S5
Fig. 7.7 The process of the proposed inter-cell interference coordination method based on
spectrum sensing technology
If the sub-band is used by the cell center users, the low power level is allocated.
If the sub-band is used by cell edge users, the high power level is allocated.
According to the power resource allocation rules, the cell center users may be
disturbed by the cell edge users from adjacent cells, which results the reduced
inter-cell interference.
7.3.2
Interference Analysis
In actual, an arbitrary sub-band k in cell m may be allocated to a user j to use, then,

in the target cell, the achievable signal to interference ratio of user i when using
sub-band k can be calculated as:
154
SIR
dm k
P
P0 k L0 k

P
m j2Slow Pinner k Lj;0 k
m j2Shigh Pedge k Lj;0 k
7:1
Where, P0 k is the transmit power of the user who has occupied the sub-band
k in the target cell. L0 k is the path loss of the user who has occupied the sub-band
k in the target cell. Slow and Shigh represent the sub-bands with low power level and
high power level respectively. m j 2 Slow denotes that sub-band k is allocated to
the center user j of cell m. m j 2 Shigh denotes that sub-band k is allocated to the
edge user j of cell m. Lj;0 k denotes the path loss of user j to the target cell.

dm k
1; sub-band k has been used by user in cell m

:
0; others
From Sect. 7.2.1, in the proposed scheme, the interference of the center users in
cell 1 mainly comes from the edge users in cell 2 and cell 3. When the center
service load in cell 1 is less than that of full load with 1/3, the co-channel interference of the center users in cell 1 mainly comes from the edge users in cell 2,
which will be reduced further when the edge service load in cell 2 is little.
The interference of the edge users in cell 1 mainly comes from the center users in
cell 2 and cell 3. As shown in Fig. 7.2, if the sub-band k is allocated to the edge user
UE1,A in cell 1 and to the center user UE2,A in cell 2 meanwhile, we tag the
interference level of UE2,A to UE1,A as auplink in the uplink, and adownlink in the
downlink. If the sub-band k is allocated to the center user UE1,A in cell 1 and to the
center user UE2,B in cell 2 meanwhile, we tag the interference level of UE2,B to
UE1,A as buplink in the uplink, and bdownlink in the downlink. We can get
adownlink Pinner L2 Pinner d0n
7:2
auplink Pinner LUE2;A Pinner d1n
7:3
bdownlink Pinner L2 Pinner d0n
7:4
buplink Pinner LUE2;B Pinner d2n
7:5
where, Pinner is the lower transmission power for the center users in cell 2. d0 is the
distance from the base station of cell 2 to the user of UE1,A in cell 1, d1 is
the distance from the user of UE2,A of cell 2 to the user of UE1,A in cell 1, d2 is the
distance from the user of UE2,B of cell 2 to the user of UE1,A in cell 1, n is
the path-loss factor.
Obviously, in the downlink, the equation of adownlink bdownlink is established.
While in the uplink, since d1 \d2 , the inequality of adownlink bdownlink is established, i.e. as to UE1,A, UE2,A is a relatively strong interference user, while UE2,B is
relatively weak. Therefore, the co-channel interference in the uplink can be effectively reduced by the proposed scheme.
7.3.3
155
Detection Algorithm
From Sects. 7.2, 7.3.1 and 7.3.2, we can see that the proposed scheme is mainly
used to improve the cell edge users performance when the service loads are full in
the traditional SFR scheme, as shown in the step of S402 in Fig. 7.7. In the
proposed scheme, when all the spectrum resources have been allocated, if a new
user lauches a service application, it will detect the spectrum with the weakest
signal to communicate through the spectrum sensing technology so as to reduce the
interference to other users. In this paper, we utilize the energy detection method and
choose the available spectrum by detecting the signal strength, i.e. the signal to
interference ratio of the received signal which can be calculated as (7.1). The best
spectrum can be obtained by the iterative method as follows:
1. For all of the sub-band N, calculate the signal to interference ratio of the
received signal on the sub-band i, the vector representation is:
SIR fSIR1 ; SIR2 ; . . .; SIRN g;
2. Dene the accuracy e, and let i = 1, j = 1;
3. If SIRi SIRj , let j = j + 1, continue, when j = N, stopotherwise, let i = i + 1,
continue;
Through the above steps, we can get the spectrum resource with the weakest
signal strength.
7.3.4
Simulation Results
In order to verify the efciency of the proposed scheme, the computer simulations
are given as follows, in which, the system bandwidth is 10 MHz. According to
3GPP TR25.814, the maximum transmit power in the uplink is 23 dBm, and cells
other conguration parameters are shown in Table 7.1.
Figure 7.8 shows the signal to interference ratio which the cell center users can
achieve versus the transmit power ratio of the cell center users to the edge users.
Table 7.1 Simulation parameters
Parameters
Value
The system bandwidth

The carrier frequency
The number of sub-band
The number of sub-carrier in each sub-band
The cells radius
The number of sub-band allocated to each user
The maximum transmit power of the mobile node
Propagation mode
10 MHz
2 GHz
27
48
1 km
1
23 dBm
128.1 + 37.6 log10(r), r (km)

the signal to interference ratio which the cell center
users can achieved
156
30
SFR
Proposed
25
20
15
10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
the transmit power ratio of the cell center users to the edge users
Fig. 7.8 The signal to interference ratio which the cell center users can achieved versus the
transmit power ratio of the cell center users to the edge users
The ordinate is evaluated by dB. It can be seen that, the interference of the cell
center users can be effectively reduced by the proposed scheme, compared with the
SFR scheme. With the increase of the transmit power ratio, the signal to interference ratio which the cell center users can achieved increases.
Figure 7.9 shows the signal to interference ratio which the cell edge users can
achieve versus the transmit power ratio of the cell center users to the edge users. It
can be seen that, the interference of the cell edge users can be effectively reduced by
the proposed scheme, compared with the SFR scheme. However, with the increase
of the transmit power ratio, the signal to interference ratio which the cell edge users
can achieved decreases.
Figure 7.10 shows the average signal to interference ratio of the cell users versus
the transmit power ratio of the cell center users to the edge users. It can be seen that,
the total interference of the cell users can be effectively reduced by the proposed
scheme, compared with the SFR scheme. With the increase of the transmit power
ratio, the signal to interference ratio which the cell users can achieved increases.
Figure 7.11 shows the average throughput of the cell users versus the transmit
power ratio of the cell center users to the edge users. It can be seen that, the average
user throughput can be effectively improved by the proposed scheme, compared
with the SFR scheme, which is due to the decrease of the interference.
the signal to interference ratio which the cell edge

users can achieved
157
30
SFR
Proposed
25
20
15
10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Fig. 7.9 The signal to interference ratio which the cell edge users can achieved versus the transmit
power ratio of the cell center users to the edge users
the average signal to interference ratio which

the cell users can achieved
30
SFR
Proposed
25
20
15
10
5
0
-5
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Fig. 7.10 The average signal to interference ratio which the cell users can achieved versus the
transmit power ratio of the cell center users to the edge users

the average throughput of the cell users(Mbps)
158
7
SFR
Proposed
1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Fig. 7.11 The average throughput of the cell users versus the transmit power ratio of the cell
center users to the edge users
7.4
Summary
In this chapter, we have studied an inter-cell interference coordination method by

spectrum sensing technology, where, the terminal with cognitive function can
automatically detect the available spectrum when requesting for service, and select
the idle spectrum or the occupied spectrum with the weakest signal strength to
realize the actual weakest interference communication, make the reduced inter-cell
interference and improved cell edge users performance. It is noted that we just
propose the idea of realizing inter cell interference coordination by spectrum
sensing. With the development of the V-MIMO technology, we can utilize the
spectrum sensing technology described in the above chapter to detect the available
spectrum, which need to be in-depth studied in the future.
References
1. Guo CL, Feng CY, Zeng ZM (2010) Cognitive radio network technologies and application.
Publishing House of Electronics Industry
2. Cheng HB, Yao YD (2010) Cognitive-relay-based intercell interference cancellation in cellular
systems. IEEE Trans Veh Technol 59(4):19011909
3. Young JC, Shin KG (2011) Opportunistic access of tv spectrum using cognitive-radio-enabled
cellular networks. IEEE Trans Veh Technol 60(8):38533864
4. Shen J (2004) 3GPP long term evolution: principle and system design. Beijing, The Peoples
Posts and Telecommunications Press
5. GPP TR 36.913 v.0.1.0. Requirements for further advancements for E-UTRA
References
159
6. Astely D, Dahlman E, Furuskar A et al (2009) LTE: the evolution of mobile broadband. IEEE
Commun Mag 47(4):4451
7. Tachwali Y, Basma F, Refai HH (2010) cognitive radio architecture for rapidly deployable
heterogeneous wireless networks. IEEE Trans Consum Electron 56(3):14261432
8. GPP TR 36.902 V9.3.0. Evolved universal terrestrial radio access network (E-UTRAN);
Self-conguring and self-optimizing network (SON) use cases and solutions
9. Peng T (2010) Investigation on hybrid networks based on cognitive radio technologies.
Beijing University of Posts and Telecommunications, Beijing
10. Ma Y, Kim DI, Wu ZQ (2010) Optimization of OFDMA-based cellular cognitive radio
networks. IEEE Trans Commun 58(8):22652276
11. Choi KW, Hossain E, KimD I (2011) Downlink subchannel and power allocation in multi-cell
OFDMA cognitive radio networks. IEEE Trans Wireless Commun 10(7):22592271
12. Sawahashi M, Kishiyama Y, Morimoto A et al (2010) Coordinated multipoint
transmission/reception techniques for LTE-advanced [Coordinated and Distributed MIMO].
IEEE Wirel Commun 17(3):2634
13. Gbor F, Chrysostomos K, Andrs R et al (2009) Intercell interference coordination in
OFDMA networks and in the 3GPP long term evolution system. J. C. 4(7):445453
14. GPP technical specication group radio access network, physical layer aspects for evolved
UTRA (2005). 3GPP TR25.814 v0.2.1
15. GPP technical specication group radio access network, requirements for evolved UTRA
(E-UTRA) and Evolved UTRAN (E-UTRAN), 3GPP, Tech. Rep. TR 25.913, Mar. 2006
16. Huawei (2005) R1-050841, further analysis of soft frequency reuse scheme, 3GPP TSG RAN
WG1#42. London, UK
17. Huawei (2005) R1-050507, soft frequency reuse scheme for UTRAN LTE, 3GPP TSG RAN
WG1 Meeting #41. Athens, Greece
18. Huawei and RITT (2006) R1-060229, Some clarications on soft frequency reuse scheme,
3GPP, TSG-RAN WG1 Ad Hoc on LTE
19. Heyman J (2006) Intercell interference management in an OFDM-based Downlink, Report
20. Ekstrom H, Furuskar A, Karlsson J et al (2006) Technical solutions for the 3G long-term
evolution. IEEE Commun Mag 44(3):3845
21. Siemens (2005) R1-050738, interference mitigationconsiderations and results on frequency
reuse, TSG-RAN WG1 Meeting #42. London, UK
22. Siemens (2005) R1-050599, interference mitigationconsiderations and results on frequency
reuse, TSG-RAN WG1 Ad Hoc on LTE, Sophia Antipolis, France
23. Siemens (2006) R1-060135, interference mitigation by partial frequency reuse, TSG-RAN
WG1 Ad Hoc Meeting on LTE, Helsinki, Finland
24. Ericsson (2005) R1-050764, inter-cell interference handling for E-UTRA, TSG-RAN WG1
#42. London, UK
25. Electronics LG (2005) R1-050833, interference mitigation in evolved UTRA/UTRAN, 3GPP
TSG RAN WG1 #42. England, London
26. Alcatel (2005) R1-050695, interference coordination for evolved UTRA uplink access, 3GPP
TSG RAN WG1 #42. London, Great Britain
27. Nokia (2005) R1-050813, UL interference control considerations, 3GPP TSG-RAN WG1
Meeting #42. London, UK
28. Instruments Texas (2005) R1-051059, inter-cell interference mitigation for EUTRA, 3GPP
TSG RAN WG1. San Diego, USA
29. Nokia (2006) R1-060291, OFDMA downlink inter-cell interference mitigation, 3GPP,
TSG-RAN WG1 #44
30. R1-050896 (2005) Description and simulations of interference management technique for
ofdma based e-utra downlink evaluation, Qualcomm Europe
31. Instruments Texas (2006) R1-060368, Performance of inter-cell interference mitigation with
semi-static frequency planning for EUTRA downlink, 3GPP TSG RAN WG1#44. Denver,
USA
160
32. Ericsson (2006) R1-061374, downlink inter-cell interference co-ordination/avoidance

evaluation of frequency reuse, TSG-RAN WG1 #45. Shanghai, China
33. Ericsson (2005) R1-061378, E-UTRA bandwidth flexibility, TSG-RAN WG1 #45. Shanghai,
China
34. Li ML (2007) The research of soft frequency reuse technology and its application in LTE
system. Beijing University of Posts and Telecommunications, Beijing
35. Patent number: 200810101751.7 A semi-static interference coordination algorithm of
OFDMA systems based on virtual cell
36. Patent number: 200910076946.5 A semi-static interference coordination algorithm
37. Patent number: 201110093697.8 An scheme, system and equipment for interference
coordination
38. Rappaport Theodore S (1999) Wireless communications: principles and practice. Publishing
House of Electronics Industry, Beijing

Cognitive Wireless Networks Using The CSS Technology

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cognitive Wireless Networks Using The CSS Technology

Uploaded by

Copyright:

Available Formats

Lecture Notes in Electrical Engineering 384

Lecture Notes in Electrical Engineering

Board of Series editors

About this Series

Communication, Networks, and Information Theory

LNEE publishes authored monographs and contributed volumes which present

More information about this series at http://www.springer.com/series/7818

Meiling Li Anhong Wang

Cognitive Wireless Networks

3 The Relationship Among the Performance Parameters in CSS . . . . .

3.3 The CSS Detection Performance . . . . . . . . . . . . . .

Censoring Based CSS . . . . . . . . .

5 CSS Technology with Relay . . . . . . . . .

5.5 Proposed Adaptive CSS Scheme

with Best Relay

Based on Soft Combination. . . . . . . . . . . . . . . . . . .

7 The SS Application in ICIC . . . . . . . .

3rd Generation Partnership Project

Selective Diversity Combining

1.1 Dissertation Background

Cognitive Radio Technology

Cognitive capability relies to a large extent on the awareness of the environment

1.2 Cognitive Radio Technology

Primary base station

Transmitter detection area

Receiver detection area

Fig. 1.1 Detection area with and without interference

Spectrum sensing has been identied as a key enabling functionality to ensure

Fig. 1.2 Spectrum decision classication

1.2 Cognitive Radio Technology

Fig. 1.3 Dynamic spectrum access, classication along regulatory status

As introduced in the previous section, DSA technology is needed to achieve a better

1.2 Cognitive Radio Technology

Fig. 1.4 Cognitive ring

Spectrum Sensing Technology

Fig. 1.5 Spectrum sensing classication

1.3 Spectrum Sensing Technology

Fig. 1.6 Classication of cooperative spectrum sensing

CR base station (BS) is naturally the FC. Alternatively, in CR ad hoc networks

Fig. 1.7 Spectrum sensing method

Spectrum Sensing Method

In cooperative spectrum sensing, each SU can utilize the sensing method in

1.3 Spectrum Sensing Technology

the interference temperature on nearby primary receivers. Since primary receivers

If prior knowledge of the PU signal is unknown, the energy detection method is

Fig. 1.8 Energy detection

Matched Filter Detection

Cyclostationary detection is more robust to noise uncertainty than energy detection.

1.3 Spectrum Sensing Technology

Fig. 1.9 The hidden terminal

can not be accurately detected by single-user spectrum sensing. To address this

8. Buddhikot MM, Ryan K (2005) Spectrum management in coordinated dynamic spectrum

31. Holland O, Muck M, Martigne P et al (2007) Development of a radio enabler for

Traditional wireless networks have predominantly used direct point-to-point or

Cooperative Communication Model

Fig. 2.1 Cooperative

2.2 Cooperative Communication Model

Fig. 2.2 The introduction

Fig. 2.3 Fixed

2.2 Cooperative Communication Model

It is needed to note that the combining factors a1 and a2 should be designed to

2.2 Cooperative Communication Model

at the relay, it entails the possibility of forwarding erroneously detected signals to