Convergence performances of various adaptive filter algorithms with application to system identification

Abstract

This paper examines the performance of mean square error (MSE) using various adaptive filtering algorithms in the adaptation process of system identification over two defined communication channels. The MSE performances of the proposed algorithms are compared with that of the standard Normalized Least Mean Square (NLMS) algorithm. To enhance the adaptation performance, non-mean-square algorithms have been utilized, for instant, Least Mean Fourth (LMF) algorithm. Also, switching algorithm led to better performance. In this paper, random step size NLMS, switching between NLMS and LMF, and LMS+F mixed norm algorithms have been implemented and their performances over two defined Finite Impulse Response (FIR) channels were tested in comparison with the standard NLMS.