Moidiefied Version of Adjusted Step Size LMS Algorithm (MASSLMS) for Adaptive Linear FIR Equalizer

Abstract

In this paper a Modified version of Adjusted Step Size Least Mean Square algorithm (MASSLMS) is proposed which overcome and avoid one of the drawback of the standard LMS and our previous proposed algorithm Adjusted Step Size Least Mean Square algorithm (ASSLMS). This drawback is the requirement of a statistical knowledge of the input signal prior to the starting training of the algorithm which is necessary to determine the fixed value of the maximum step size (i.e. the upper bound value) in the initialization stage of the ASSLMS algorithm. In this proposed algorithm an appropriate time varying value of the maximum step size was calculated based on inversely proportional of the instantaneous energy of the input signal vector. Then this time varying upper bound value of the step size is used to guarantee the stability of adjusted step size of the algorithm which is a recursively adjusted based on rough estimate of the performance surface gradient square . The proposed algorithm does not need trial and error for choosing the value of the maximum step size (µMAX) compared with ASSLMS and standard LMS algorithms. The proposed algorithm shows through computer simulation results faster and low level of miss-adjustment in the steady state compared with LMS and ASSLMS for three different types of channel in adaptive linear equalizer system.