ADAPTIVE

INSTRUCTOR :DR. AAMER IQBAL BHATTI

FILTERS I
Text Book: Adaptive Signal Processing
B. Widrow and S. D. Stearns, Prentice
Hall, Englewood, 1985.
ISBN: 0130040290
Reference Book : Adaptive Filter Theory
(4th Edition)
by Simon S. Haykin,
ISBN:0130901261
Definition

 Adapt, v.t., 1. To make suitable to requirements or

conditions; adjust or modify fittingly (Random House
Dictionary, 1971)
Course Contents

 Basic adaptive filters

 Optimization Concepts
 Least mean square filters
 Other adaptive algorithms
 Application to Real World Problems
 Stochastic processes and Models
 Prediction theory and Linear Prediction
Course Material 4
 Lecture PDFs
 One assignment per two week
 Lab exercises in MATLAB and SIMULINK
 Files will be available at yahoo group
 Address of the group is adaptivefilters@groups.yahoo.com
 Submit your email addresses (preferably at yahoo) to Salman
 Labs will be arranged according to class strength
Concept 5

 An adaptive entity is a system whose structure is alterable or

adjustable in such a way that its behaviour or performance
improves through contact with its environment, according to
some desired criterion.
Everyday Example 6

 The input signal intensity to a radio receiver varies a lot.

 Need to keep the input power level with in certain range.
 Requires an online tuning of signal level
 A variable gain is the solution
 This is the concept of Automatic Gain Control (AGC)
AGC 7

 AGC is a multiplier whose gain is inversely proportional to

the signal level.
 Thus trying to maintain the input power range.
 AGC acts as an adaptive gain to the receiver system.
Purpose of this course 8

 Introduce and motivate adaptive filters

 Derive the underlying laws and equations
 Explore probable areas of applications
 Implementation of adaptive filters in such applications
Adaptive Filters ??? 9

 Resembles with ordinary FIR filters

 However taps/coefficients/parameters of the filters are NOT
designed
 These taps are adjusted online to meet certain criterion
 The change in taps is done through clever adaptation
algorithms
Characteristics 10

 Automatic adaptation
 Specific filtering
 Do not need to be designed
 Self-healing
 More complex
Applications Overview 11

 Modeling and system identification

 Design and testing
 Adaptive noise cancellation
 Echo cancellation
 Inverse modeling
 Inverse filtering
 Deconvolution
 Equalization
Properties 12

 Time-varying
 Nonlinear
 Principle of superposition does not hold
 Fixed filters
 Insufficient system information
 Time varying systems/channels
 Optimal for a certain class of systems/channels
Types of Adaptation 13

 Open loop adaptation

 Only input/system signal knowledge is used
 Relatively straightforward
 Supervisory kind of control
 System specific knowledge is required
 Cannot cope with time drifts
 14

Input Signal Processor

Output
Signal

Adaptation algorithm
Other
Data
Types of Adaptation 15

 Closed loop adaptation

 Output signal is used for filter adjustment
 Automatic adjustment
 Responsive to changing conditions
 16

Input Signal Processor Output Signal

Adaptation algorithm

Performance Calculation
Other Data
Terminology 17
 Estimator or a filter is a system that is designed to extract information
about a prescribed quantity of interest from noisy data
 There must be some criteria to distinguish between the useful signal and
the noisy data
 Filters in frequency domain like low pass high pass etc. use frequency as
the basis to distinguish between the two
Example 18

 Transmitted signal in a communication system is

distorted by the
a) Inter-Symbol Interference due to non ideal behavior of
the channel
b) Noise from the environment
 Job of the receiver is to estimate the transmitted signal
Example 19
Kinds of Estimation 20

 Filtering: is an operation that incurs the extraction of information about a

quantity of interest at time n
Kinds of Estimation 21

 Smoothing: is a posteriori form of estimation . In that data measured after

the time of interest are used in the estimation
Kinds of Estimation 22

Prediction: is the forecasting side of estimation

Linear Optimum Filters 23

 Statistical approach
 Availability of the following information is assumed
a) Statistical characteristics of the noisy data
b) Statistical characteristics of useful signals
Linear Adaptive Filters 24

 For Stationary input we mean:

Input is generated by the process whose statistical characteristics are not
changing with time
 For stationary inputs the resulting solution is commonly known as Wiener
Filter
 For non stationary environment Kalman Filter is the suitable solution
wiener Filter and Motive for 25
Adaptive Filter
 Design of a wiener filter requires a prior information about
the statistics of the data to be processed.
 wiener filter is optimum if the exact parameters are known
 If the parameters are changing with time or the exact values
are not known than wiener solution is not optimum
Wiener Filter and Motive for 26
Adaptive Filter
 An adaptive filter solves the problem by ‘estimate and plug
‘ approach
 This solution is costly in terms of computational resources it
requires
 To mitigate this problem we use Adaptive Filters
Performance Evaluation of the 27
Adaptive

Filter
Performance of an Adaptive Filter algorithm is analyzed based on
following criteria
a) Rate of convergence
b) Miss adjustment
c) Tracking
d) Robustness
e) Computational requirements
f) Structure
Challenges for Adaptive Filter 28
Designer
 Understand the capabilities and limitations of various
adaptive filters
 Understanding the selection of the appropriate algorithm for
the application at hand
Applications 29

 Identification: Adaptive filter is used to provide a linear model that

represents the best fit to an unknown plant
Applications 30

 Inverse Modeling: provide an inverse model for the process generating

noisy data
Applications 31

 Prediction: function of an adaptive filter to provide the best estimate of the

future values of a signal
Applications 32

 Interference: Cancellation: adaptive filter is used to cancel

unknown interference contained in a primary signal
Task 33

 Read Chapter 1 of Widrow’s book

 Read the “Introduction and Background” from the Haykin’s
book (optional)