Abstract:

In order to solve the contradictory effect between convergence speed and excess mean square error of the Least Mean Square (LMS) algorithm,some variable step size (VSS) LMS algorithms are proposed.Among these algorithms,the E-LMS algorithm associates the step size with mean square error while the R-LMS algorithm associates the step size with error correlation function.The E-LMS algorithm has poor performance under low signal noise ratio (SNR) condition,and R-LMS algorithm has poor tracking performance for burst systems.In this paper,a modified algorithm based on error correlation function is proposed,whose tracking performance is enhanced by the mean square error.The computer simulation results show that the algorithm has good performance of anti-noise and tracking.

Key words: least mean square;variable step size;error correlation function;mean square error