Variable step size modified clipped LMS algorithm

Abstract

In this paper we introduce an Modified Clipped LMS (MCLMS) algorithm with a variable step size. In the MCLMS algorithm two parameters, the step size and the threshold control the convergence rate of the adaptive filter coefficients and also determine the final mean-square error. The computational complexity decreased dramatically by a large threshold. However, this selection results in a low convergence rate. Since the convergence time is inversely proportional to the step size, a large step size is often selected for fast convergence. But a large step size results in an increased final mean square error. Therefore in this paper we choose a large threshold and propose a variable step size for the MCLMS algorithm. The advantages of this proposed variable step size and a large threshold selection are that the computation complexity is low, final mean square error is low and that the convergence is fast.