Crisscross Optimization Algorithm for the Designing of Quadrature Mirror Filter Bank

During designing of two channel Quadrature Mirror Filter(QMF) bank, there occurs many errors such as peak reconstruction error, error in pass band, error in stop band, error in transition band and amplitude distortion. There are many meta-heuristic algorithms for the minimization of various errors of two channel QMF banks but they have drawbacks that either their convergence speed is slow or they have poor solution accuracy or premature convergence. In this paper, a concept of the Crisscross Optimization method for minimizing designing error of two channel linear phase QMF bank in frequency domain is introduced. This problem is taken as single objective function. The crisscross optimization algorithm makes use of two important operators that are horizontal crossover (HCX) and vertical crossover (VCX). The HCX operator works by dividing the problem space into simple hypercube with smaller search probability while VCX operator makes a way out for the stagnant dimensions which may be indifferentiable and cause premature convergence. The simulation results are compared with the modified particle swarm optimization, artificial bee colony, XSG and Levenberg Marquart method. From the comparison it is obvious that convergence speed and good solution accuracy.