Sie sind auf Seite 1von 40

DESIGN OF FIR FILTER BY USING

OPTIMISATION TECHNIQUES
INTRODUCTION

 FIR filters are more preffered due to its


guaranteed stability, free from phase distortion, low
coefficient sensitiviy,simplicity etc..

 Most advanced form of designing optimum filter by


using optimisation algorithms

 Initially PM method is used

2
INTRODUCTION…

 Other optimization algorithms are evolutionary,


swarm optimization algorithms, cat optimization
algorithms etc…

 Filter coefficients are find out using these algorithms.

 Most desirable filter with minimum error is selected.

Title of the Project 3


LITERATURE REVIEW

 Design of linear phase FIR low pass filter using


particle swarm optimization algorithm.
 Design and optimization of FIR filter using genetic
algorithm.
 Design of FIR digital filter using differential
evolution algorithm.
 Better results using simple iterations.
 Compare characteristics like pass band ripples, stop
band attenuation etc.
 Less execution of time.

4
DIGITAL FILTER

 Designing of an optimum filter has prime importance in


present era.

 Here optimization algorithms used for designing.

 The program has to be iterated many times to obtain


better results.

 Main aim to minimize errors.

Title of the Project 5


DIGITAL FILTER…

This is the characteristic equation of a digital filter

Where,
M is order of filter
h(n) is the filter impulse response coefficient
Filter has (m+1) number of coefficients.

Title of the Project 6


DIGITAL FILTER…

E(W)-Error function

𝐻𝑖(𝑒𝑗𝑤) –Ideal filter response.

𝐻𝑑(𝑒𝑗𝑤)-Actual filter response.

G (w) -Weighting function.

Title of the Project 7


OPTIMIZATION TECHNIQUES

 Different optimisation techniques used

1.Genetic algorithm.

2.DE algorithm.

3.Particle swarm optimization algorithm


.
4.Adaptive DEPSO.

5.Cat swarm optimisation.

Title of the Project 8


GENETIC ALGORITHM

 Based on the concept of survival of fittest

 Basic steps:
A. Population
set of coefficients are initial population.

B.Evaluation
measurement of fitness value.

Title of the Project 9


GENETIC ALGORITHM…
C.Selection
selection of best fit individual.
D.Reproduction
two fittest chromosomes selected to produce off springs.
E.Cross over.
Exchange of genes between selected individual.
D.Mutation
Random changes introduce genetic diversity in
population
Repeat the steps until desired optimization coefficients

Title of the Project 10


FLOWCHART OF GA

Title of the Project 11


DE ALGORITHM

.
 Based on principles of GA

 Crossover and mutation operation work directly on


continuous valued vectors.

 DE algorithm relays on mutation operation.

 Steps
• generate an initial population of solutions.

• Calculate fitness for initial population.

Title of the Project 12


DE ALGORITHM…..

• Select three solutions at random for each parent.


• A mutant vector produced for each target vector.
• Parent vector mixed with mutant vector gives trial vector.
• Performance of trial vector and parent vectors compared.
• Repeat until number of offspring equals population size
• Parent is replaced by fittest offspring.
 ADVANTAGES
• stable than GA.
• Applicable to real valued problems despite noisy,multi
modal spaces

Title of the Project 13


FLOW CHART OF DE

Title of the Project 14


PSO ALGORITHM
 Developed through simulation of bird flocking.

 First introduced by kennedy and Eberhart

 Each particle/bird tries to modifies its position.

 Distance between current position and g best/p best is


evaluated.

 Velocity modifies towards this value.

 Best suitable value selected.

 Accurate ,flexible and less chance of frailer

Title of the Project 15


FLOW CHART OF PSO

Title of the Project 16


ADAPTIVE DEPSO

 Hybrid version of DE evolution and PSO.

 Used fitness based adaptive cross over rate.

 Better fitness value calculated by DE algorithm

 Best value is transferred to the trial vector.

 Most robust and accurate results.

Title of the Project 17


FLOWCHART OF ADEPSO
start

Initial population
.

Evaluate fitness using DE

Fffittest individual towards trial vector

Initiate particle with position and velocity

Eval;uateevaluate fitness

Update pbest /gbest particle

cconverg
ence

stop
Title of the Project 18
CAT SWARM OPTIMIZATION ALGORITHM

 Imitates natural behaviour of cat.


 Two characteristics of cat
• Resting with slow movement.-seeking mode.
• Chasing with high speed.-tracking mode.
 These characteristics combined to solve optimisation
problems.
 More effective in reducing stop band ripple factor.

Title of the Project 19


. CAT SWARM OPTIMIZATION ALGORITHM..
STEPS
 Generate random population.
 Randomly select a point of reference as a cat.
 Generate all possible candidates of a cat.
 Find the fitness value of candidates.
 If fitness value is better change the position of cat.
 Calculate the velocities of each dimension of cat and
change the resultant position.
 If this value is better optimize the position of cat
Title of the Project 20
FLOWCHART OF CSO

Title of the Project 21


CALCULATED FILTER COEFFICIENTS USING
DIFFERENT ALGORITHM

Title of the Project 22


CALCULATED FILTER COEFFICIENTS USING DIFFERENT
ALGORITHMS…

 Simulation results performed in MATLAB R2014a

 Order N is taken as 20.

 Number of coefficients is 21.

 Frequency response of filter designed from


coefficients.

 Main objective function to find out pass band ripples

Stop band ripples and stop band attenuation


Title of the Project 23
FREQUENCY RESPONSE PLOT DERIVED USING
MATLAB 2014A

Title of the Project 24


COMPARATIVE STUDY OF DIFFERENT
ALGORITHMS

Title of the Project 25


COMPARATIVE STUDY OF DIFFERENT
ALGORITHMS…

 Pass band ripple is least in case of PM.

 Effective reduction in stop band attenuation is in CSO


and ADEPSO.

• Signal stopped in stop band.

• Leakage noise will be less.

 Better results occurs in CSO and ADEPSO algorithms.

Title of the Project 26


PLOTS DERIVED IN MATLAB 2014(OWNWORK)
PLOTS DERIVED USING RGA

Title of the Project 27


PLOTS DERIVED USING RGA

Title of the Project 28


PLOTS DERIVED USING DE

Title of the Project 29


PLOTS DERIVED USING DE…

Title of the Project 30


PLOTS DERIVED USING PSO

Title of the Project 31


PLOTS DERIVED USING PSO…

Title of the Project 32


PLOTS DERIVED USING ADEPSO

Title of the Project 33


PLOTS DERIVED USING ADEPSO…

Title of the Project 34


PLOTS DERIVED USING CSO

Title of the Project 35


PLOTS DERIVED USING CSO…

Title of the Project 36


INFORMATION OF FILTER USING DIFFERENT
OPTIMIZATION ALGORITHM

Title of the Project 37


COMPARISON OF PLOTS

• Filters designed using RGA algorithm is not a perfect


linear phase filter
• Group delay and phase delay plots of RGA are not
constant.
• Step response and impulse response plots using
different algorithms are almost same.
• Poles located at z=0 in all pole - zero plots .ie poles
located inside unit circle.So all the filters are stable

Title of the Project 38


CONCLUSION
 PSO is more effective than GA and DE.
 More enhanced results obtained using hybrid
algorithms.
 Latest approach of designing filter is ADEPSO,CSO
algorithms.
 Most accurate and stable results using ADEPSO and
CSO.
 CSO is selected as most efficient algorithm.

39
REFERENCES
1.N.karaboga and B cetinkaya”design of digital fir filters using DE
algorithm". circuit system signal processing,vol.25.pp.649-660,jan
2006

2.Smandal,S.PN ghoshal”design of optimal linear phase FIR filter using


improved particle swarm optimization” ,I international journal on
signals and image processing,vol.03,pp.5-9,jan 2012.

3.N.karaboga ,”a new design method based on artificial bee colony


algorithm for digital FIR filters:. journal of the Franklin
institute,vol.346,4,pp.328-348, march 2009.

4.S. Mukherjee, R. Kar, D. Mdandal, S. Mondal, and S. P. Ghoshal,


“Linear Phase Low Pass FIR Filter Design Using Improved Particle
Swarm Optimization”, IEEE Student Conference on Research and
Development, Durgapur, India,Dec 2011, pp. 358-363.
40

Das könnte Ihnen auch gefallen