Modified Normalized Least Mean Square Algorithm with Improved Minimization Criterion

Abstract

In this paper we develop an improved minimization criterion for normalized least mean squares (NLMS) algorithm using past weight vectors and adaptive learning rate. The proposed criterion minimizes the summation of each squared Euclidean norm of difference between the currently updated weight vector and past weight vector. The result of the modified NLMS algorithm has lower misalignment than the conventional NLMS algorithm for various SNR. The simulation shows that the convergence rate of proposed NLMS algorithm is faster as the previous weight vectors and SNR increases.