Proportionate M-estimate adaptive filtering algorithms: Insights and improvements

Abstract

In the literature, the proportionate least mean M-estimate (PLMM) algorithm exhibits good performance when dealing with sparse systems in the presence of impulsive noises. In this paper, the mean and mean-square behaviors of three recursion types of the PLMM algorithm are studied in depth. We derive analytically the stability, transient and steady-state results of these PLMM recursions, and find that they can achieve the same performance when properly choosing the step-size and the proportionate matrix. To improve the filter performance in both convergence rate and steady-state error, we derive a variable step-size (VSS) scheme and then present the VSS-based PLMM algorithms. In addition, based on the adaptive decorrelation strategy aiming at the colored input signals, the VSS adaptive decorrelation PLMM algorithms are developed to further speed up the convergence. Computer simulations have verified our theoretical analyses and the effectiveness of the proposed algorithms.