FIR FILTER DESIGN USING MULTIOBJECTIVE PARTICLE SWARM OPTIMIZATION

Abstract

Abstract The proposed method designs a FIR filter with arbitrary frequency response and reduced delay. The coefficients of this FIR filter are real and the delay is less than half of the filter coefficients. The FIR filter in this approach has non linear group delay. The filter is designed using multi-objective approach minimizing certain objective functions that are responsible for generating a filter of lower delay and magnitude response of unity in the passband. The multiobjective constraints are tailored by incorporating an evolutionary algorithm with multiobjective approach. In this proposed approach Particle Swarm Optimization with multiobjective optimization(MOPSO) produces a set of non-dominated solutions called pareto optimals. The MOPSO takes a set of real coefficients of the FIR as the population and using multiobjective error formulation of amplitude response and group delay gives optimal FIR filters. The error formulation for magnitude response is to have a response of 1 in passband and 0 in stopband, and for the group delay, it must be as close to linearity as possible in passband. These solutions can tailor all types of requirements of the decision maker. The proposed approach has been compared with weighed least square method and the experimental results have shown that the magnitude response and delay characteristic using proposed approach are better than those achieved by weighed least square approach.