0% found this document useful (0 votes)
84 views26 pages

EE501 Adaptive Filter Design: Instructor: Dr. Farhan Khalid

This document provides an outline for an adaptive filter design course. The course will cover background topics in discrete-time signal processing and linear algebra. It will then cover discrete-time random processes, optimum filters including Wiener filters, and various adaptive filter algorithms including LMS, NLMS, and RLS. The course will also cover adaptive IIR and FIR filters, adaptive arrays, and applications of adaptive filtering. Key reference books for the course are also listed.

Uploaded by

Sana Saad
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
84 views26 pages

EE501 Adaptive Filter Design: Instructor: Dr. Farhan Khalid

This document provides an outline for an adaptive filter design course. The course will cover background topics in discrete-time signal processing and linear algebra. It will then cover discrete-time random processes, optimum filters including Wiener filters, and various adaptive filter algorithms including LMS, NLMS, and RLS. The course will also cover adaptive IIR and FIR filters, adaptive arrays, and applications of adaptive filtering. Key reference books for the course are also listed.

Uploaded by

Sana Saad
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 26

EE501 Adaptive Filter Design

Instructor: Dr. Farhan Khalid

1
Course Outline
• Background:
– Discrete-Time Signal Processing
– Linear Algebra
• Discrete-Time Random Processes
• Optimum Filters
– FIR and IIR Wiener Filter
• FIR Adaptive Filters
– Steepest Descent Adaptive Filter
– LMS, NLMS, LLMS and Variable Step-Size Algorithms
– Applications
• Adaptive Recursive (IIR) Filters
– IIR LMS Adaptive Filter
• Recursive Least Squares Algorithms
• Adaptive Arrays and Adaptive Beamforming

2
Books
• Monson H. Hayes, Statistical Digital Signal
Processing and Modeling, Wiley
• Bernard Widrow, Samuel D. Stearns, Adaptive
Signal Processing, Prentice-Hall
• Dimitris G. Manolakis, Vinay K. Ingle, Stephen
M. Kogon, Statistical and Adaptive Signal
Processing, Artech House

3
Adaptive Filtering

4
Background

5
Discrete-Time Signals
• Discrete-time signals
x (n )

-2 -1 1 3 4 6 n

x (n )  xa (nTs )

x (n )   x (k ) (n  k )
k 
6
Discrete-Time Signals
• Unit sample
1 ; n0
(n )  
0 ; otherwise

• Unit step
1 ; n0
u(n )  
0 ; otherwise
• Relation between unit step and unit sample
n
u(n )  
k 
(k )
7
Discrete-Time Signals
• Complex exponential (periodic) signal
e jn 0
 cos(n 0 ) + j sin(n 0 )

0 a real constant

• Right-sided sequences: equal to zero for n  0


Example: unit step

8
Discrete-Time Signals
• Left-sided sequences, example: Time-reversed
and delayed unit step
1 ; n  n0
x (n )  u(n 0  n )  
0 ; n  n0

• Two-sided sequence: neither right-sided nor


left-sided
Example: infinite-length complex exponential
9
Discrete-Time Systems
x (n ) y(n )  T x (n )
T

• Linearity and Shift-Invariance


• Causality
• Stability
• Invertibility

10
Time-Domain Descriptions of LSI Filters
• Difference Equation for IIR Filter:
q p
y(n )   b(k )x (n  k )   a(k )y(n  k )
k 0 k 1

• Difference Equation for FIR Filter:


q
y(n )   b(k )x (n  k )
k 0

• FIR Filter Impulse Response


q
h(n )   b(k ) (n  k )
k 0 11
Discrete-Time Fourier Transform (DTFT)
• DTFT of x (n ) is a complex-valued function of
the continuous (frequency) variable,

X (e ) 
j
 x (n )e
n 
 jn

• A sufficient condition for uniform convergence


of the sum to a continuous function of is


n 
x (n )  

i.e., x (n ) must be absolutely summable. 12


Discrete-Time Fourier Transform (DTFT)
 jn
• Since e is periodic in with a period of 2 ,
j
X (e ) is also periodic with a period of 2
j
• X (e ) is conjugate symmetric if x (n ) real
* j
X (e )  X (e )
j

• DTFT in polar form (in terms of magnitude and


phase)
X (e )  X (e ) e
j j j x( )

• For real signals


j
X (e )  X (e
j
) and x ( )   x( )
13
Discrete-Time Fourier Transform (DTFT)
• Frequency Response: DTFT of the unit sample
response of an LSI system

H (e ) 
j
 h(n )e
n 
 jn

• DTFT defines how a complex exponential is


changed in amplitude and phase by the
system
• Inverse DTFT (IDTFT):
1
x (n )  
j jn
X (e )e d
2  14
Discrete-Time Fourier Transform (DTFT)
• Convolution in time domain
y(n )  x (n ) * h(n )
• equals product of the transforms in frequency
domain
Y (e )  X (e )H (e )
j j j

• Parseval’s Theorem: sum of the squares of a


signal equals intergral of the square of its DTFT

1

2 2
x (n )  
j
X (e ) d
n  2  15
The z-Transform
• z-Transform of a discrete-time signal

X (z )   x (n )z
n 
n
where z  re jw

• z-Transform becomes the DTFT when r  1 or


z e jw


X (e )  X (z ) z e j 
jw
 x (n )e
n 
 jnw

16
The z-Transform
• Region of Convergence(RoC): defines those
values of z for which the sum converges
• Finite length sequence: RoC includes all values
of z (except possibly z  0 or z  )
• Right-sided sequences: RoC is exterior of a circle
z  R where R is a positive number
• Left-sided sequences: RoC is interior of a circle
z  R+ where R+ is a positive number
• Two-sided sequences: RoC is an annulus
R  z  R+ 17
The z-Transform
• For a conjugate symmetric sequence
*
x (n )  x (n )
its z-transform satisfies the relation
* *
X (z )  X (1 / z )

• This symmetry condition is useful in dealing


with the power spectrum.

18
The z-Transform
• System Function: z-transform of the unit sample
response

H (z )  
n 
h(n )z n

• FIR Filters:
q q
H (z )   b(k )z k
 b(0) (1  z k z ) 1

k 0 k 1 zeros

19
The z-Transform
• IIR Filters: q
q

 b(k )z k
 (1  z
k 1
k
1
z )
H (z )  k 0
p
 b(0) p
zeros

1 +  a(k )z k
 (1  p k
1
z )
k 1 k 1 poles

• All-pole Filter:
b(0) b(0)
H (z )  p
 p
1 +  a(k )z k  (1  pk z 1
)
k 1 k 1 poles 20
The z-Transform
• Real-valued coefficients a(k) and b(k)
(equivalently, real-valued h(n)):
* *
H (z )  H (z )

and the poles and zeros of H(z) will occur in


conjugate pairs

21
Special Classes of Filters
• Linear phase filter
• Allpass filter: constant magnitude frequency
response
• Minimum phase filter: all poles and zeros
inside unit circle

Reading: Section 2.2.6

22
Filter Flowgraphs

23
Filter Flowgraphs

24
Filter Flowgraphs

Reading: Section 2.2.7

25
Discrete Fourier Transform (DFT)
• N-point DFT
N 1
X (k )   x (n )e
n 0
 j 2 kn /N

• Inverse DFT (IDFT)


1 N 1
x (n )  
N k 0
X (k )e j 2 kn /N

• Fast Fourier Transform (FFT) algorithms: faster


(reduced-complexity) evaluation of DFT
Reading: Section 2.2.8
26

You might also like