The variable step-size wavelet transform-domain LMS adaptive filter algorithm

Abstract

In this paper, the wavelet transform domain least mean squares (WTDLMS) adaptive algorithm with variablestep-size (VSS) is established. The step-size changes according to the largest decrease in mean square deviation. To keep the computational complexity low, the Haar wavelet transform (HWT) is utilized as a transform. In addition, the mean square performance analysis of the VSS-WTDLMS is studied in the stationary and nonstationary environments and the theoretical relations for transient and steady-state performances are established. The simulation results show that the proposed VSS-WTDLMS has faster convergence rate and lower misadjustment than conventional WTDLMS. The theoretical relations are also verified by presenting various experimental results.