Hybrid genetic algorithm for the design of modified frequency-response masking filters in a discrete space

Abstract

This paper presents the design of high-speed, arbitrarybandwidth sharp finite impulse response filters with signed powers-of-two coefficients based on a modified frequency-response masking (FRM) structure. A novel hybrid genetic algorithm (HGA) is proposed to jointly optimize all subfilters in a discrete space. The proposed HGA introduces the simulated annealing technique into the genetic algorithm (GA) optimization process and effectively prevents the GA from prematurely converging. It is shown, by means of examples, that FRM filters designed by the HGA achieve a significant reductionin the number of bits. © Birkhauser Boston 2006.