An Efficient Method for Designing Multiplier-Less Non-uniform Filter Bank Based on Hybrid Method Using CSE Technique

Abstract

This paper presents an efficient method for optimal design of multiplierless non-uniform tree-structured filter banks using hybrid evolutionary algorithm. Phenomenal reduction in number of adders is achieved by applying common subexpression elimination (CSE) on canonic signed digits (CSD) and binary represented filter coefficients. The proposed methodology utilizes the proficiencies of particle swarm optimization and artificial bee colony algorithms to achieve magnitude response of a prototype filter for non-uniform filter bank (NUFB) of 0.7071 at quadrature frequency. Single optimization technique is employed for designing the computationally efficient, continuous and quantized coefficients of a prototype filter for NUFB. A comprehensive study on CSE technique such as vertical, horizontal and mixed CSE has been also made for designing NUFB. The simulation results reflect that the proposed method shows better performance in terms of adder gain and reconstruction error. The hybrid method yields 33 % of adder gain with CSD technique and 55 % adder gain in case of CSE technique. B I. Sharma

Optimization of synthesis filters for hybrid filter bank DACs

“…Several filter match design methods such as linear programming [22], second-order cone programming [23] have been proposed to eliminate the aliasing error and minimize the distortion. Most works [24,25,26,27,28,29] focus on designing the digital filter to match a pre-set analog filter bank while the parameter of these analog filters are not discussed. Once an analog filter bank is chosen, the system performance would be determined using a specific design method.…”