A Study on Normalized LMS Algorithm Using Refined Filtering Technique

We investigate the convergence behavior of the normalized least mean square (NLMS) algorithm in the structure of a linear transversal filter. At the n-th iteration, the traditional NLMS transversal filter generates the n-th output signal by using linear convolution of the n-th input vector and the n-th coefficient vector. Based on this result, the n-th coefficient vector is updated to the n + 1-th coefficient vector. We attempt a refined filtering (RF) approach to the NLMS transversal filter, to generate another output signal by linear convolution of the n-th input vector and the n+ 1-th coefficient vector. Theoretical analysis and computer simulation demonstrate the effectiveness of the RF technique. Key-Words: linear transversal filter, normalized LMS algorithm, adaptive line enhancer, refined filtering