一种变步长LMS自适应滤波算法

A LMS Adaptive Filtering Algorithm with Variable Step Size

本文基于对最速下降（算）法梯度的优化估值，提出了线性－指数（Ｌ．Ｅ）自适应滤波算法。文中指出：对最速下降法的梯度优化估值和对误差信号非线性处理，是改善ＬＭＳ算法性能的等同措施。文中的理论分析与计算机模拟结果都证实了Ｌ．Ｅ算法的性能优于［１］所提出的线性－符号（Ｌ．Ｓ）算法。理论分析与计算结果吻合较好。

In this paper,the Linear-Exponential(L.E) algorithm are presented that is bassed on the optimal gradient estimation of steepest descent method .This point out that in order to improve the LMS algorithm performance, the gradient are optimally estimated, or the error signal are nonlinearity processed,these methods are equipollent. The theoreticaloanalysis and computer simulation show the performance of L.E algorithm is better than the L. S algorithm performance that is presented by [1]. The computer simulation result shows that very good match with the theoretical analysis.