You are on page 1of 12

Wireless Pers Commun

DOI 10.1007/s11277-017-4071-0

Interference Analysis and Mitigation Techniques


in Wireless Body Area Networks

Pradnya H. Ghare1 A. G. Kothari1

 Springer Science+Business Media New York 2017

Abstract Wireless body area network (WBAN) is a network of RF enabled wearable and
implantable sensor nodes. Due to use of these networks in healthcare applications, relia-
bility is an important metric. However, due to coexistence of other wireless technologies as
well as presence of many WBANs in the same area, interference can persist. This inter-
ference can degrade reliability. In this regard, this paper presents the analysis of the
interference issue in body area networks. Also this paper has made an effort to suggest an
algorithm for interference mitigation using data rate parameter and thereby trying to
improve the interference scenario in terms of packet loss ratio in WBANs.

Keywords Wireless body area network  Interference  Interference mitigation 


Healthcare  Packet loss

1 Introduction

With advances in sensor technology, there is a large increase in use of wearable devices. A
survey in [1] has predicted a tremendous increase up to 480 million in the number of
wearable devices till 2018. Wireless body area network is a network of wearable devices
worn by human being for measurement of physiological parameters. These networks
operate with Industrial, Scientific and Medical (ISM) band of frequencies. Due to use of
ISM frequency band, other wireless devices are expected to coexist with WBAN. This
coexistence can create interference which could hamper reliability of these networks.
Considering various applications of BAN in public places such as hospitals, markets,
public transport systems, marathons and first responders system [2], it appears that there is
a large possibility of more than one BAN in the vicinity of each other. Due to mobile

& Pradnya H. Ghare


ghareph22@gmail.com
1
VNIT, Nagpur, India

123
P. H. Ghare, A. G. Kothari

nature of WBAN, there is a possibility of co-location and co-channel interference. Due to


use of unlicensed frequency band such as ISM (Inductrial, Scientific and Medical), two
BANs can concurrently transmit the signals using the same frequency channel. When all
the sensors broadcast the signals, there is a possibility that main signal of interest may
superimpose with the interfering signal from other BANs. This is shown in the Fig. 1.
Wireless link between a sensor node and coordinator is also affected by rapid movements
of body parts which causes performance degradation of the desired signal. This also causes
packet collision and energy waste which affects WBAN QoS. In this regard, it is important
to know about impact of various parameters for generation of interference. This paper has
tried to show the impact of various parameters on signal quality for multiple coexisting
WBAN environments.
As WBANs are used in healthcare applications, loss of data will be dangerous partic-
ularly for critical medical applications and the purpose of use of WBAN for healthcare
applications will no longer persist. Therefore it is very important to mitigate this inter-
ference in order to be assured of a reliable data communication. This paper has tried to
suggest priority based adaptive data rate mechanism for interference mitigation in coex-
isting BANs.
This paper is an extended version of [3].
Following section presents related work. Section 3 discusses about factors contributing
to signal quality degradation. Interference effects observed and suggested interference
mitigation mechanism is presented in Sect. 4. Finally Sect. 5 concludes the results of the
work.

2 Related Work

Interference analysis in wireless body area networks has assumed lot of significance in
current healthcare scenario. Some of the mitigation methods suggested by different
researchers includes different approaches like use of game theory, scheduling and
switching approaches etc. [4, 5]. Game theory based approaches rely on negotiation
between BANs to reach a Nash equilibrium. However, in high interference scenario such
negotiations cannot take place practically. A mitigation mechanism based on the perfor-
mance of modified and modulated hermite pulses (MHP) for narrowband interference
mitigation in the 4.944.99 GHz has been proposed in [6]. Another interference mitigation
method is based on variable data rates, duty cycles and modulation schemes [7]. Here in
[7], the metric of choice to measure interference and improve the interference scenario has

Fig. 1 Interference in WBAN

123
Interference Analysis and Mitigation Techniques in Wireless...

been SINR. The proposed algorithm leads to better SINR when compared to its unmiti-
gated scenario. In [8], a time hopping code assignment strategy is used to improve Body
Area Network performance and a strategy to increase the lifetime of the network is also
introduced. A distributed TDMA-based beacon interval shifting scheme is suggested in [9]
to reduce beacon collision in BANs. This is achieved through carrier sensing before beacon
transmission.

3 Factors Responsible for Generation of Interference

The various sources for generation of interference in WBAN arise from Physical as well as
MAC layer parameters. Some of them to mention are mobility, placement of nodes,
shadowing, path loss, channel frequency, access mechanism at MAC layer etc. A pictorial
representation of various factors contributing to interference is shown in Fig. 2.

3.1 Quantitative Measurement of Inter-User Interference

To quantitatively measure the signal degradation due to Inter-User Interference, packet


error rate (PER) is estimated. Packet error rate (PER) is the ratio of number of wrongly
received data packets to the total number of transmitted packets. This is a function of bit
error rate which in turn depends on modulation scheme and Signal to Interference Noise
Ratio. The interference can be represented by following equation.

Fig. 2 Factors responsible for interference generation

123
P. H. Ghare, A. G. Kothari

X
I Sj tPr Di;j t 1
j6i

Pr(Di;j (t)) is the power received from sensors of coexisting BANs to coordinator of main
BAN which represents interference and Sj (t) represents presence or absence of coexisting
BAN. Pr(Di;j (t)) is dependent on many factors such as path loss, distance, frequency, path
loss exponent, fading etc. All these factors are responsible for generation of interference in
WBAN. According to Friss model [10], path loss at a distance d can be represented by

PLd Gda 2

where G is System Loss Constant, a refers to Path Loss Exponent and d is the distance
between transmitter and receiver. The concurrent data transmissions of BANs cause
interference to the neighboring BANs. At a given time instant t, the transmission status of
neighboring BAN is denoted by Sj(t):Sj(t) 1 if the transmitting node is active otherwise
0. Thus SINR of BAN i is given by,
Pr Di t
SINRi t P 3
S
j6i j tPr Di;j t Pn

Here, Di (t) is the distance between Coordinator and Sensor of main BAN of Interest i.
Di;j (t) is the distance between coordinator of BAN i and sensor node of BAN j at time t and
Pn is the Background Noise Power. Assuming Offset Quadrature Phase Shift Keying
(OQPSK) to be the modulation scheme being utilized, the average Bit Error Rate is given
by
r
1 2Eb
BERi erfc 4
2 No
Z tmax Z xmax p
1
BERi erfc SINRi t PXdxdt 5
tmin xmin 2

P(X) is the Probability Distribution function. Subsequently, the Packet Error Rate is given
as follows.
mkm2  
X mk rm2 mkrm2
PERi m BERi 1  BERi
6
r1
r2

Here, m kr m=2 represents a binomial coefficient with (r m=2) combinations


out of a set of (m k) elements. From Eq. (6), the PER of a BAN can be calculated. Thus
PER will vary according to number of interfering BANs, distance between these BANs,
path loss exponent, fading and modulation scheme.
Following section shows simulation results using Matlab for packet error rate.
Parameter and their values under Considerations:-
Since Log-Normal Distribution is the best possible fit for almost all of the BSN
scenarios [11], P(x) is determined using Log-Normal fading model.
The mean l and Standard deviation r values in evaluating BER varies according to the
physical position and location of sensor where it is implanted on BAN user [12]. So
l 0:75; r 0:80 is considered. It is the optimum value because all the possible real
time values of r and l are centered on this value.

123
Interference Analysis and Mitigation Techniques in Wireless...

BANs are power hungry and the range of transmittable powers available in BANs are
from [-25, and 0 dBm].
In any typical BAN scenario, the value of PLE roughly ranges from 2 to 4. As the
attenuation of signals from the same BAN is according to on-body path loss model and
that from Interfering signals is according to the Inter-Body path loss model, the path
loss exponent values were taken as 3.38 for on-body and (23) for Inter-Body links
[13].

3.2 PER Versus Distance of Interferers

Here the term interfering distances refer to the average of distances of Interfering BANs
(IBANs) in meters. It is obvious from the Fig. 3 that closer the IBANs, larger will be the
PER at main BAN. From the Fig. 3, it can be deduced that, acceptable level of PER  0:1
can be obtained for distances greater than 3 m. However in BAN applications such as
waiting hall of a hospital, managing the distance between BANs is not possible. Hence if
density of BANs within a small area is more, PER will be above threshold  0:1.
The effect of transmission powers of Interfering BANs is shown in Fig. 4.

3.3 PER Versus Number of Interfering BANs

For different values of transmission powers, PER is plotted against number of interferers as
shown in Fig. 5. These values were observed keeping the average Inter-BAN distance as 3
m. tp refers to transmission power of main BAN sensors. When tp is more, PER is
relatively less at main BAN. It means that more Packet Loss will be there at the main BAN
if the power of IBANs dominate that of main BAN sensors. Mainly, this plot tells what
should be the transmission power of main BAN sensors so that more number of BAN users
can be accommodated in given vicinity.

Fig. 3 Variation of PER as a function of distance of interferers

123
P. H. Ghare, A. G. Kothari

Fig. 4 Variation of PER as a function of transmission power

Fig. 5 Dependence of PER on the number of interfering BANs

3.4 PER Versus Path Loss Exponent (PLE)

It can be seen from Fig. 6 that with an increase of Inter-body PLE there is a significant
amount of decrease in the PER. This decrease in PER is because with more Inter-body
PLE, more will be the attenuation of Interfering signal before interfering the main BAN
signals and subsequently a relatively weak signal interferes with the main BAN. This leads
to less packet drop, finally less PER. An Inter-body PLE greater than 2.6 leads to a
negligible interference with the main BAN signals. All the above individual plots
demonstrate the variation of PER with respect to each of the parameters. Combining the
above results gives an idea of the trade-off among these parameters in maintaining a
minimum packet error rate allowing a reliable data transmission.

123
Interference Analysis and Mitigation Techniques in Wireless...

Fig. 6 Impact of PLE on the PER values

3.5 Observing the Effects of Interference Using File Transfer Model

In the previous section, the quantitative analysis of interference in terms of PER has been shown. While
the statistics are comprehensive, it does not show the extent of the effect of the interference in real time
files, which can be sent by sensors to the sink node, for example ECG signal of a patient. To investigate
this particular case, a scenario using four sensor nodes and a cluster head (in 4  1 arrangement)
which represent a cluster each was created in Netsim simulator. ECG signal files are sent by
the sensors to the sink node through the same channel (i.e same center frequency). The signal
files, after being received by the sink node, has been reconstructed there to get back the sent
file. A thorough comparison of the sent file from the sensor and the reconstructed file at the
sink node will depict the extent of effect of interference in this scenario.
Figure 7 depicts the sent ECG signal data file from one particular sensor to the sink
node, projected in 7 bit ASCII format and converted to hexadecimal representation. At the
sink node or server, the packet loss ratio (PLR) in this scenario is approximately 9.09%.
Due to this, a certain number of packets have been dropped.
The sent file is captured at the sink node and reconstructed there, which is shown in
Fig. 8. It can be argued that the difference between sent and reconstructed file is obviously
due to the packet loss due to unmitigated interference.

Fig. 7 Sent ECG file from sensor to sink node

123
P. H. Ghare, A. G. Kothari

4 Adaptive Data Rate Based Interference Mitigation

Adaptive Data Rate, usually denote the varying data rate of a particular device depending
upon fulfillment of certain criteria. In this case, data rate of a particular sensor changes
depending on number of packet drops as well as priority of sensor node.
As WBANs are used for monitoring different physiological parameters such as body
temperature, heart rate, emotion status etc, each one has a different priority. For instance,
ECG or EEG sensors seems to have more important information than sensors indicating body
temperature or blood glucose level and hence a WBAN will be consisting of a set of high
priority and low priority sensor devices. The process of variation of data rate is carried out in
runtime and the adjustment is done automatically by the sensors itself. The criteria upon
which these adjustments are done are exclusively user defined so that it can be changed based
upon application, environmental parameters or user expectations from the system.
To determine when, how and which data rate is going to be used in the Run Time
Adaptive Data Rate mechanism, an algorithm is used. It is given below.
Let M represents number of data rate levels, D represents the step size in data rate used
in kbps, Rmax represents maximum data rate and Rmin is the minimum data rate, d
represents a miniscule amount, u represents current packet loss ratio(PLR) and u represents
target packet loss ratio.
Algorithm: Run Time Adaptive Data Rate
Required: M data rates, Priority type
Required: u (target Packet Loss Ratio)
Initial: u (current Packet Loss Ratio) = 0;
Initial: Rmax R1 (default Data Rate);
1. R R1
2. D Rmax  Rmin=M;
3. If u  udandR [ Rmin; then
4. if (priority=High) then
5. R R1D;
6. else
7. R R1  2D;
8. end if;
9. end if;
10. Send Packet;

Fig. 8 Reconstructed ECG file at sink node

123
Interference Analysis and Mitigation Techniques in Wireless...

11. Calculate u;
12. If (Packets != Null), then
13. Go to Step 3;
14. End.
Considering the different healthcare parameters such as ECG,EEG and EMG etc, the range
of the data rate values (e.g. from 250 to 100 Kbps) have been chosen in this work.Hence as
per the above algorithm, 3 different data rates have been chosen to switch upon, namely
200, 150 and 100 Kbps other than the default data rate of 250 Kbps. So, in this case, M
comes to be three. It is worth noting that M, u, u, Rmax, Rmin are all changeable or user
defined and can be set in an application specific manner.
The values of data rate levels (M) and target packet loss ratio are dependent on par-
ticular healthcare application, required precision in terms of mitigation and also on
hardware and software constraints of sensor nodes. Initially, the number of data rate levels
(M) and target packet loss ratio u is required from the user. The current PLR (u) is set as 0
and maximum data rate (Rmax) is equated to default data rate. The step size is calculated
based on Rmax, Rmin and M. As long as the current PLR is lesser than target PLR, packets
are sent at default data rate and the resultant current PLR is calculated subsequently during
run-time. Now, if this PLR overhauls the target PLR, then the data rate is adjusted by
decreasing it by the amount of step size D as long as it is greater than pre-defined minimum
data rate. The algorithm ends when the number of packets to be transmitted by a device
reaches zero.
Based on the sensor wise statistic of unmitigated scenario and priority of sensor nodes,
the data rates and the corresponding number of packet drops (representative of PLR) for
each test case is chosen. The next step in this mechanism is to optimize the packet drop
ratio along with the data rate even more, so that the solution covers higher degrees of
packet drops and thereby reducing the PLR. This two-step strategy is followed in every
case mentioned, where mitigation mechanism is applied. As we move from Case 1 to Case
5, the packet drop limits for switching to a lower data rate has been made gradually more
stringent. In this way, we can arrive at a more robust solution to achieve a better PLR and
also compare one particular scheme with respect to its previous one in terms of PLR (as
shown in Fig. 9). In this work, test cases to procure observations are obtained using the
algorithm. The values of the data rates are grouped along with different target PLRs
(represented in absolute number of packets dropped) for the sake of comprehensive
observations.
Adaptive Power Level based interference mitigation: The interference analysis section
indicates dependence of transmission power of sensor nodes on packet loss. Use of
adaptive transmission power mechanism can reduce packet loss. The rationale behind the
adaptive power level mechanism is that if a sensor node which is dropping packets
switches its transmission power from one particular level to another one, the corresponding
interfering nodes will cease to interfere. For example, in the given scenario in this work
any sensor node can switch from its basic transmitting power level of 1020 or 100 mW
subject to fulfillment of certain packet loss criteria. In this mechanism, the power levels at
which the sensors will switch to for transmitting packets are to be decided by the user. The
values of these power levels will have to be chosen after considering certain constraints.
These constraints include the permissible Specific Absorption rate (SAR) values for Body
Area Applications, the initial energy and power handling capabilities of the sensors.

123
P. H. Ghare, A. G. Kothari

Fig. 9 Statistics of interference mitigation techniques

4.1 Results

4.1.1 Parameters Used in Simulation

As discussed in previous section five different test cases are simulated using the suggested
algorithm.
Simulation environment created in Netsim is as follows.
An area of 25  20 sqm is chosen, consisting of 6 WBANs corresponding to six patients.
Each WBAN consists of 9 sensors, including one sensor cum cluster head and 8 other
sensors to measure vital parameters.
A sink node (equivalent to a server application enabled device) is placed in the same
environment which receives all the data from the sensor node.
The default data rate of the uplink from sensor to cluster head and then to sink node is
250 Kbps.The channel is centered at ISM band frequencies (2.405 GHz). Initially
performance results for single BAN without using any mitigation techniques have been
shown in Table 1. This is followed by results of mitigation techniques for the same.
Five different test cases for observation is made by grouping various data rates with
packet drop criteria of any sensor node (representing PLR) to examine the results of
mitigated scenario. The respective cases are,
Case 1. Default Scenario (250 Kbps data rate throughout)
Case 2. Adaptive Data Rate-I (200 Kbps after 10 Packet Drops, 150 kbps after 30 packet
drops by a sensor)
Case 3. Adaptive Data Rate-II (150 Kbps after 8 Packet Drops, 100 Kbps after 15 packet
drops by a sensor)
Case 4. Adaptive Data Rate-III (150 Kbps after 5 Packet Drops, 100 Kbps after 10
packet drops by a sensor)
Case 5. Adaptive Data Rate and Adaptive Power Level (150 Kbps and 20 mw of
transmitter power after 5 Packet Drops, 100 Kbps and 100 mw of transmitter power after
10 Packet Drops by a sensor)

Table 1 Packet delivery statistics of unmitigated scenario


Packet delivery statistics at sink node

Number of packets transmitted PLR System runtime(Sec)


10,530 3:86% 100

123
Interference Analysis and Mitigation Techniques in Wireless...

Comparative effect of the mechanisms on interference can be deduced from the Fig. 9. It
can be seen that Case2, Case 3 and Case 5 has improved the PLR from its previous scenario,
but on the contrary in Case 4, PLR has deteriorated. So, it is clear that an optimization of data
rate and threshold metric values are required after extensive implementation, experimenta-
tion and observation. It can be seen that Case 2 is the first bit of interference mitigation
mechanism applied on the scenario. So, the improvement of Case-2 (14.77, and 0.56% of
PLR) over its previous case is highest when compared to improvement of other cases over its
immediate previous one. In Case 3, a marginally different and evidently improved mecha-
nism is used. So, essentially it is nothing but an optimization technique. The improvement
upon its immediate preceding case here is 9.85, and 0.23% of PLR. Then, in Case 4, the
mechanism of Case 3 has been altered to optimize further and it does not improve the Packet
Loss Ratio. So, after a certain point, any more change in threshold metric will result in
deterioration, as can be observed from figure. In this particular case, the deterioration of
Packet Loss Ratio is 8.5, and 0.17% of total number of packets sent towards sink node. In Case
5, exclusively the effect of the Run Time Adaptive Power Level scheme has been shown,
because while migrating from Case 4 to Case 5 the data rate related metrics are kept exactly
the same. This mechanism improves the interference rate by nearly 10 percent, 0.30% of the
PLR. These two mechanisms, Run Time Adaptive Data Rate and Run Time Adaptive Power
Level can be used simultaneously to reduce the packet loss rate, by more than 26 percent
(Almost 1% of PLR) when compared to normal unmitigated scenario.

5 Conclusion

This paper addresses the interference issue in WBAN via quantitative analysis and also
demonstrates simulation based results. WBAN is considered here for healthcare applica-
tions. Analysis shows that number of BANs, transmission power of sensors as well as path
loss exponent have a great impact on level of interference.It appears that for distances upto
2 m, interference level is very high even for few number of coexisting BANs.
Run Time Adaptive Data Rate along with a Run Time Adaptive Power Level mitigation
mechanism has been used for interference mitigation. The mitigation techniques, when
used both separately and simultaneously, improve the Packet Loss Ratio and hence the
interference. It is expected that the combination of common manual calibration techniques
and the mitigation mechanisms would be able to eradicate Inter WBAN interference to a
large extent.In the future, suggested mitigation technique will be tested on actual hardware.

References
1. www.ABIResearch.com.
2. Batalin, M. (2011). PHASER: Physiological Health Assessment System for Emergency Responders. In
1st workshop on body area network technology and applications (pp. 18).
3. Moitra, S, Ghare, P. H., & Kothari, A. G. (2014). Interference analysis and mitigation techniques in
wireless body area network. In The 18th international symposium on WIRELESS PERSONAL MUL-
TIMEDIA COMMUNICATIONS (WPMC), Special session on Combedded systems.
4. de Silva, B., Natarajan, A., & Motani, M. (2009). Inter-user interference in body sensor networks:
Preliminary investigation and an infrastructure based solution. In Sixth international workshop on
wearable and implantable body sensor networks (p. 3540).

123
P. H. Ghare, A. G. Kothari

5. Wu, G., Ren, J., Xia, F., Yao, L., & Xu, Z. (2010). DISG: Decentralized inter-user interference
suppression in body sensor nerworks with non-cooperative game. In International conference on
autonomic and trusted computing (pp. 256261).
6. Rout, D. K., & Das, S. (2014). Interference mitigation in wireless body area networks using modified
and modulated mhp. Wireless Personal Communications, 77, 13431361.
7. Yang, W. B., & Sayrafian-Pour, K. (2012). Interference mitigation using adaptive schemes in body area
networks. International Journal of Wireless information Networks, 19, 193200.
8. Domenicali, D., Nardis, L. D., & Benedetto, M. G. D. (2009). UWB body area network coexistence by
interference mitigation. In International conference on ultra-wideband (ICUWB 2009).
9. Kim, S., Kim, Jin-Woo, & Eom, Doo-Seop. (2012). A beacon interval shifting scheme for interference
mitigation in body area networks. Sensors, 12, 1093010946.
10. Balanis, C. A. (1982). Antenna theory, analysis and design (1st ed.). NewYork, NY: Harper and Row.
11. Sun, W., Ge, Y., & Wong, W.-C. (2012). A light-weight inter-user interference mitigation method in
body sensor networks. In 8th IEEE international conference on wireless and mobile computing, net-
working and communications (pp. 3440).
12. Reusens, E., Joseph, W., Latre, B., Braem, B., Vermeeren, G., Tanghe, E., et al. (2009). Characteri-
zation of on-body communication channel and energy efficient topology design for wireless body area
networks. IEEE Transactions on Information Technology in Biomedicine, 13, 933945.
13. Smith, D. B., Miniutti, D., Lamahewa, T. A., & Hanlen, L. W. (2013). Propogation models for body-
area networks: A survey and new outlook. IEEE Antennas and Propogation Magazines, 55(5), 97117.

Pradnya H. Ghare is currently working as Assistant professor in


Electronics and Communication Engineering Department of Visves-
varaya National Institute of Technology, Nagpur, India. Her research
interests are wireless sensor networks, body area networks. She has 10
years of teaching experience and has authored around 08 publications.

A. G. Kothari is currently working as Associate Professor in Elec-


tronics and Communication Engineering Department of Visvesvaraya
National Institute of Technology, Nagpur, India. He is also one of the
coordinators for Center of Excellence of COMMBEDDED SYS-
TEMS: Hybridization of Communications and Embedded Systems
under World Bank assisted project of TEQIP 1.2.1. He received his
Ph.D. in 2010 from VNIT, Nagpur. Rough Sets, Communication and
antennas are his field of interests. Applications of Rough Set based
signal processing are his special interest. He has authored around 25
publications and has contributed book chapters for reputed publica-
tions. He has 20 years of experience in teaching.

123

You might also like