DIGITAL SIGNAL PROCESSING

Muhammad Majid ([email protected]) Digital Signal Processing

 Lecture 2: Summary
• Discrete Time Systems
• Linear Time Invariant System
• Linear Time Invariant System Examples
• Properties of LTI System
• Stable and Causal LTI Systems

Muhammad Majid ([email protected]) Digital Signal Processing

 Lecture 3: Outline
• Linear Constant Coefficient Difference Equation
• Eigenfunction of LTI System
• Frequency Response of LTI System
• Discrete Time Fourier Transform
• Course Project

Muhammad Majid ([email protected]) Digital Signal Processing

 Linear Constant-Coefficient
Difference Equations
• For all computationally realizable LTI systems, the
input and output satisfy a difference equation of the
form
N M

å a y [ n - k ] = åb x [ n - k ]
k k
k=0 k=0

• This leads to the recurrence formula which can be

used to compute the “present” output from the
present and M past values of the input and N past
values of the output

Muhammad Majid ([email protected]) Digital Signal Processing

 Linear Constant-Coefficient
Difference Equations
• First Order Example
• Consider the difference equation
• y[n] =ay[n-1] +x[n]
• We can represent this system by the following
block diagram:

Muhammad Majid ([email protected]) Digital Signal Processing

 Linear Constant-Coefficient
Difference Equations
y[n] =ay[n-1] +x[n]
Consider input x[n]=Kδ[n] and Initial value condition
y[-1]=c
•For n>-1

Muhammad Majid ([email protected]) Digital Signal Processing

 Linear Constant-Coefficient
Difference Equations
• For n>=0

• For n<-1 difference equation can be written as

y[n] =ay[n-1] +x[n]
ay[n-1] = y[n] - x[n]
y[n-1] = a-1(y[n] –x[n])
y[n] = a-1(y[n+1] –x[n+1])
• Using auxiliary condition y[-1]=c, we can compute
y[n] for n<-1

Muhammad Majid ([email protected]) Digital Signal Processing

 Linear Constant-Coefficient
Difference Equations
• For n<-1

• Overall results of recursive computation

Muhammad Majid ([email protected]) Digital Signal Processing

 Linear Constant-Coefficient
Difference Equations

• We computed output both in positive and negative

direction, beginning with n=-1
• Clearly procedure is noncausal
• When K=0 means x[n]=0 but y[n] = an+1c
• A linear system must require y[n]=0 if x[n]=0
• This system is not linear
• If the input is shifted by n0 samples x1[n] = Kδ[n-n0]
then output y1[n] = an+1c + Kan-n0u[n-n0]
• The system is not time invariant
Muhammad Majid ([email protected]) Digital Signal Processing
 Linear Constant-Coefficient
Difference Equations
• An important class of LTI systems of the form
N M

å a y [ n - k ] = åb x [ n - k ]
k k
k=0 k=0

• The output is not uniquely specified for a given input

• The initial conditions are required
• Linearity, time invariance, and causality depend on
the initial conditions
• If initial conditions are assumed to be zero system
is linear, time invariant, and causal

Muhammad Majid ([email protected]) Digital Signal Processing

 Eigenfunction of LTI System

• Consider an input to the LTI system

x[n]=ejωn
• then output of the LTI system is convolution sum
y[n]=x[n]*h[n]
¥
y[n] = å x [ k ]h[n - k]
k=-¥
¥
y[n] = å h [ k ]x[n - k]
k=-¥

Muhammad Majid ([email protected]) Digital Signal Processing

 Eigenfunction of LTI System

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Frequency response H(ej) is the complex function of
frequency and can be represented in terms of real
and imaginary parts as

• Or in terms of magnitude and phase

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
Example: Ideal Delay System

y[n] = x[n - nd ]
• Consider an input x[n]=ejωn then

• Frequency response of ideal delay system is

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Alternative method to obtain frequency response is to
use impulse response of ideal delay system i.e.,
h[n] = δ[n-nd]

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Frequency response of discrete time LTI system is
periodic with period 2π
¥
H (e jw
) = å h [ n]e - jw n

n=-¥

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Frequency response of discrete time LTI system is
periodic with period 2π

• In general

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Frequency response of ideal lowpass filter

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Frequency response of ideal highpass filter

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Frequency response of ideal bandstop filter

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Frequency response of ideal bandpass filter

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
Example: Moving Average System

• The impulse response of moving average system is

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System
• Frequency response of moving average system is

Muhammad Majid ([email protected]) Digital Signal Processing

 Frequency Response of LTI
System

M1=0
M2=4

Muhammad Majid ([email protected]) Digital Signal Processing

 Discrete Time Fourier
Transform
• Many sequences can be expressed as a weighted
sum of complex exponentials as
xn 
1 
2  
X e j
 
e jn
d (inverse transform)

• Where the weighting is determined as

    xne

j  jn
Xe (forward transform)
n  

• X(ej) is the Fourier spectrum of the sequence x[n]

• It specifies the magnitude and phase of the
sequence
• The phase wraps at 2 hence is not uniquely
specified
Muhammad Majid ([email protected]) Digital Signal Processing
 Discrete Time Fourier
Transform Pair
    xne 1 
 

Xe j  jn
and xn   X e j
e jn
d
n   2  

• Let’s show that they constitute a transform pair

• Substitute first equation into second to get
1     jm  jn

 1  jn m 
x̂n   
  xme e d   xm  e d
2  m  
 
 m    2  

 sinn  m
1  jn m   0 for n  m
 n  m
2  
e d 
 1 for n  m
 n  m

x̂n   xmn  m  xn
m  

Muhammad Majid ([email protected]) Digital Signal Processing

 Existence of Discrete Time
Fourier Transform
• For a given sequence the DTFT exist if the infinite
sum convergence
    xne

j  jn
Xe
• Or n  

 
X e j   for all 

    xne
  
Xe j

n  
 jn
  xn e
n  
 jn
  xn  
n  

• So the DTFT exists if a given sequence is absolute

summable
• All stable systems are absolute summable and have
DTFTs
Muhammad Majid ([email protected]) Digital Signal Processing
 Absolute and Square
Summability
• Absolute summability is sufficient condition for DTFT
• Some sequences may not be absolute summable but
only square summable


 xn
2

n  

• To represent square summable sequences with DTFT

• We can relax the uniform convergence condition
• Convergence is in mean-squared sense
Xe    xne
 M
j  jn
X M ( e ) = å x [ n]e- jw n
jw

n   n=-M

lim  X e
M 
   X e 
j
M
j
2
0


Muhammad Majid ([email protected]) Digital Signal Processing

 Absolute and Square
Summability
• To represent square summable sequences with DTFT
• We can relax the uniform convergence condition
• Convergence is in mean-squared sense

    xne
 M
Xe j  jn
XM (e jw
)= å x [ n]e- jw n
n   n=-M

lim  X e
M 
   X e 
j
M
j
2
0


• Error does not converge to zero for every value of 

• The mean-squared value of the error over all 
does

Muhammad Majid ([email protected]) Digital Signal Processing

 Example: Ideal Lowpass
Filter
• The periodic DTFT of the ideal lowpass filter is

1   c
 
Hlp e j

0 c    


• Find the inverse

sin wc n
hlp [ n ] =
pn

• Not causal
• Not absolute summable but it has a DTFT?
• The DTFT converges in the mean-squared sense
• Role of Gibbs phenomenon
Muhammad Majid ([email protected]) Digital Signal Processing
 Example: Ideal Lowpass
Filter

Muhammad Majid ([email protected]) Digital Signal Processing

 Course Project
Proposal
• Group of maximum of three students

• Submit project proposal by Thursday 06-10-2022

• At most one page (excluding references)
• What is the problem?
• What digital signal processing techniques do you
plan to experiment with?

• Project proposal presentation 3 minutes on Friday

07-10-2022

Muhammad Majid ([email protected]) Digital Signal Processing

 Course Project
Final Report and Presentation
• Submit by Friday before Final Exam
• Project Report at most 8 pages (excluding
references)
• Project Presentation 15 minutes Q/A 5 minutes
• Report Structure
• Introduction
• What is the problem?
• Why is it an important problem?
• Techniques to Tackle the Problem
• Brief review of previous work concerning this problem
• Brief description of the techniques chosen and why

Muhammad Majid ([email protected]) Digital Signal Processing

 Course Project
Final Report and Presentation
• Report Structure
• Empirical Evaluation
• Compare empirically the techniques for complexity,
performance, ease of use, etc.
• Conclusion:
• What is the best technique?
• Is any technique good enough to declare the problem
solved?
• What future research do you recommend?

Muhammad Majid ([email protected]) Digital Signal Processing