Professional Documents
Culture Documents
Processing
Introduction
Prof.Hansa Shingrakhia
Course Overview
2
Books
Texts
Digital Signal Processing: Principles, Algorithm &
Application,4th ed,Proakis, Manolakis, Pearson
Discrete Time Signal Processing:2nd Ed
Oppenheim & Schafer, PHI, 2003
3
What is DSP
Performing an operation on a signal using digital
techniques
4
Processing in DSP
Algorithm
Mathematical description of a process
Basic processes
Multiply
Convolute
Correlate
Major applications
Filtering
Transformation
5
DSP Applications in
Communications
Speech, Audio & Video Compression (Cellular
Telephone, HDTV, MPEG, MP3)
Text to speech conversion
Voice recognition
Filtering (Recover signal from noise)
Modulation & Demodulation
Encryption
Echo Cancellation
Noise Cancellation
Line Equalization
Detection and Estimation
Performance measure
6
Advantages of DSP
Flexibility
Accuracy
Storage
Economical
Limitation
Speed of operation
7
DSP
8
Signal Classification
A signal is described by a function of a
single or multiple variables
The variable(s) can be
Real valued scalar quantity s (t ) A sin t
Complex valued quantity s (t ) cos t j sin t
Vector quantity (multi-channel signal)
s(t ) s1 (t ) s2 (t ) s3 (t )
Signals generated by multiple sources or
sensors
A multi-dimensional signal is function of
more than one independent variables I ( x, y , t )
Picture (2 dimensions x,y)
9
Signal Classification
Multi-channel multi-dimensional signal
Color video with R,G,B color signals
I r ( x, y , t )
I ( x, y , t ) I g ( x, y , t )
I b ( x, y , t )
A moving point in three dimensional space
y P(x,y,z,t)
x
z
10
Other classifications
Continuous-time Discrete-time
x(t) x(n)
Discrete valued
Continuous valued
quantized
Deterministic Random
Past, present and future are
Cannot be predicted
known
Expressed only
Can be expressed
probabilistically
mathematically
11
Continuous-time sinusoidal
signal
xa(t)=Acos(t+), -<t<
=Acos(2Ft+)
Negative frequency
A/2
t + Positive frequency
t +
A/2
13
Discrete sinusoid
x(n)= Acos(n+) -<n<
= Acos(2fn+)
n is the sample number
is the frequency in radians/sample
f is the frequency in cycles/sample
=2/16
14
Properties of discrete-time
sinusoid
A discrete-time sinusoid is periodic only if its frequency f is a
rational number
ie. x(n+N)=x(n) for all n
Smallest value of N for which this relation holds is the
fundamental period
Proof:
For a sinusoid with fundamental frequency f0 to be periodic we
must have
Cos[2f0(n+N)+]= Cos[2f0n+]
2f0N=2k
Hence f0 = k/N where k, N are integers
15
Discrete-time sinusoid-
example
1 31
N=31, k=4
=8/31 rad/sample
f=4/31 cycles/sample
16
Continuous and discrete
sinusoids
Discrete-time sinusoids whose frequencies are
separated by integer multiples of 2 are identical
For continuous - time sinusoids
x1 (t ) cos(1t ), x2 (t ) cos( 2t )
If 1 2 then x1 (t ) x2 (t )
Lowpass
Lowpass
Filter
Filter Sampler
Sampler Quantizer
Quantizer Encoder
Encoder Digital
Analog Output
Input
18
Signal waveforms in ADC
Amplitude
Continuous analog signal
time
T , f F / Fs , 1 2 f 1 2
, F / T , F f .Fs
/T /T
Fs / 2 F Fs / 2
Discrete-time
21
Aliasing
22
Sampling theorem
If highest frequency in an analog signal xa(t) is Fmax=B and
the signal is sampled at fs>2Fmax=2B, then xa(t) can be
recovered from the samples using an interpolation function
sin 2Bt
g (t )
2Bt
Convolution sum
Recovered signal xa(t) is expressed as
n n
n sin 2B (t n / 2 B)
xa (t ) xa g t xa
n Fs Fs n 2 B 2B(t n / 2 B)
n
where xa xa nT x n are the samples of xa(t) and
Fs
Fs 2 B 2 Fmax is the Nyquist sampling rate
23
Reconstruction of signal
from its samples
sin 2Bt sin 2Fmax t sin Fs t sin t / T
g (t )
2Bt 2Fmax t Fs t t / T
1 g(t)
-3T -2T -T 0 T 2T 3T
xa(t)
T
24
25