一种新型变步长LMS自适应滤波算法及性能分析

A New Type of LMS Algorithm and Its Performance Analysis

针对现有LMS算法不能同时提高收敛速度及降低稳态误差的矛盾。提出一种在步长参数u(n)与误差e(n)之间建立了一种分段线性函数关系。该关系具有在误差e(n)接近零处缓慢变化的优点,克服了一些变步长LMS算法在自适应稳态阶段u(n)取值偏大的缺陷。实验证明,本算法具有计算量小,抗干扰能力强,性能可以明显改善等优点。

For existing LMS algorithm could not improve the convergence speed and lower steady-state error of conradictions at the same time,the algorithm uses a piecewise linear relationship between and the error signal when falls into different ranges.The piecewise linear relationship show a good property that is very small when approaches zero.It overcomes the shortcoming of some other algorithms where becomes large.The new method already tested in the experiment has less computational complexity and stronger robustness.