Improved Variable Step-Size Partial Update LMS Algorithm

Abstract

Partial update (PU) methods play a critical part in least-mean square (LMS) algorithms with an aim to reducing computational complexity. This paper presented a novel variable step-size PU procedure for LMS algorithms via the performance metric, called error difference indicator. With the aid of such an indicator, the PU procedure can determine the proper step size and works efficiently with a low-cost hardware implementation. Simulation results revealed that the proposed PU method speeds up the convergence rate of LMS algorithms as compared to the existing methods.