A new variable step-Size LMS algorithm based on piecewise function

Abstract

In this paper, a new improved variable step size adaptive algorithm is proposed. The step size is devised to piecewise functions for increasing the performance of the algorithm. By using the characteristics of the each stage, the proposed algorithm can improve the convergence speed and decreases the steady state error. The proposed algorithm has simple formulation and easily setting parameters. The effect of different parameters in piecewise functions are discussed. The simulation results exhibit the proposed algorithm has good tracking ability, convergence rate and small steady state error.