IIR FILTER DESIGN USING MULTI OBJECTIVE PARTICLE SWARM OPTIMIZATION

Abstract

IIR filter design is considered very important and difficult task in digital signal processing. One of the disadvantages of IIR filters is their non linear phase characteristics. Evolutionary algorithms are introduced in recent past into IIR filter design methods. IIR filter design requires concurrent minimization of order of filter, magnitude response error and linear phase response error. The proposed method designs an IIR filter with minimum order, linear phase and minimum magnitude response error. The algorithm finds the coefficients of the transfer function of desired filter. The filter is designed using Pareto based multi-objective optimization approach minimizing three objective functions simultaneously. In this thesis Multi Objective problem is solved using Pareto based Multi Objective Particle Swarm Optimization (MOPSO). Pareto based algorithms produces a set of non-dominated solutions called Pareto optimal set in one run of algorithm. It is left to the decision maker to select one solution from Pareto optimal set based on application of the filter. In literature, only Genetic based algorithms have been used in IIR filter design. PSO results in a better convergence rate. Results of proposed approach are compared with conventional approaches. Different filter types namely Low pass, High pass, Band pass and Band stop are designed. Performance of the approach depend upon some factors like number of iterations, size of repository, population size and other parameters of swarm. Experimental results show that proposed approach result in better magnitude response and have more linear phase than other approaches.