New Method of LMS Variable Step-Size Formulation for Adaptive Noise Cancellation

Abstract

Least mean square (LMS) is a widely used steepest descent algorithm known with efficient tracking ability of small mean square error (MSE) but with low convergence speed. In contract to the fixed step size, variable step size was introduced to improve the convergence speed while maintaining the minimal MSE. In this work, a new method was formulated to determine the variable step size of the LMS algorithm. Simulation results are presented to support the experimental analysis for the performance evaluation and comparison. Result reveals that the performance the of new formulated variable step size algorithm is better compare to the conventional LMS algorithm.