Finite Word Length FIR Filter Designing Using BBPSO variants

Abstract

The utilization of optimization techniques for designing digital fi lters has become widespread over the recent years. FIR lter optimization is a multi-model problem. The objective of this work is to derive an optimum set of filter coefficient s for the given speci cations for di erent type of lters. The optimization problem is phrased as least squared error (LSE) between the speci ed and designed lter response in the frequency domain. This paper presents an alternative approach for the designing of linear phase Finite impulse response (FIR) high pass lter by using Bare Bone particle swarm optimization (BBPSO) with Cauchy jump strategy technique incorporated with ripple constraint and nite word length constraint in order to achieve atter frequency response and hardware e cient lter design respectively. The inclusion of nite word length become very crucial when the lter needed to be implement on a digital processor with a xed point implementation. It is expedient to use BBPSO with jumps for optimization because it require minimum parameter tuning as in the case of other PSO variants. Two case of 20th and 30th order lters have been considered for both low pass and high pass applications. The normalized frequency response justify that proposed BBPSO based technique vanquish the other PSO variants in term of accuracy, robustness, convergence speed and stability. This work also include a case study based on ECG signal fi ltering for illustration of eff ectiveness of the designed lter.