A Variable Step-Size Proportionate NLMS Algorithm for Identification of Sparse Impulse Response

作者：Liu Ligang^*; Fukumoto Masahiro; Saiki Sachio; Zhang Shiyong

来源：IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences, 2010, E93A(1): 233-242.

DOI：10.1587/transfun.E93.A.233

• adaptive filter
• proportionate adaptation
• normalized least-mean-square
• sparse impulse response
• variable step-size

Recently, proportionate adaptive algorithms have been proposed to speed up convergence in the identification of sparse impulse response. Although they can improve convergence for sparse impulse responses, the steady-state misalignment is limited by the constant step-size parameter. In this article, based on the principle of least perturbation, we first present a derivation of normalized version of proportionate algorithms. Then by taking the disturbance signal into account, we propose a variable step-size proportionate NLMS algorithm to combine the benefits of both variable step-size algorithms and proportionate algorithms. The proposed approach can achieve fast convergence with a large step size when the identification error is large, and then considerably decrease the steady-state misalignment with a small step size after the adaptive filter reaches a certain degree of convergence. Simulation results verify the effectiveness of the proposed approach.

• 出版日期2010-1
• 单位复旦大学