VARIOUS LMS-TYPE ALGORITHMS: STUDY AND COMPARISON

Abstract

In this dissertation various LMS-Type algorithms are described. LMS adaptive filter becomes unstable, unless the adaptation constant is properly chosen. We therefore find that in choosing a suitable value for the adaptation constant, a compromise must be made between fast tracking and low excess mean-squared error. Variable step size algorithm improves the convergence. rate while sacrificing little in steady state error. These algorithms have very good convergence speed and low steady state misadjustment. Furthermore, the tracking performance of these algorithms in nonstationary environment is relatively insensitive to the choice of the parameters of the adaptive filter. Tracking performance of these algorithms is close to the best possible performance of the least mean square (LMS) algorithm, for large range of values of the step- size of the step size adaptation process. A simulation approach is presented for performance evaluation of all these algorithms, for simulation purpose a model of unknown time-variable system is considered. Simulation results are presented to support the theory and to compare the performance of these algorithms in stationary and nonstationary environment. Among these entire algorithms modified variable step size (MVSS) algorithm provides fast convergence at early stage of adaptation while ensuring small final misadjustment.