一种改进的变步长LMS自适应算法及分析

现有的变步长LMS算法中,大都采用建立步长因子与误差信号的函数关系的方法,以提高算法的收敛速度和跟踪性能,但由于未考虑输入信号对算法性能的影响,使得当输入信号发生变化后,稳态误差明显增大。为此,在现有算法的基础上,引入了输入信号因子,提出了一种改进算法。该算法可根据瞬时误差和输入信号来调整步长因子,使算法不仅能保持较高的收敛速度和跟踪性能,还可在输入信号变化的情况下,保证较小的稳态误差。理论分析及仿真实验表明新算法的性能优于现有算法。

An Improved Variable Step-size LMS Adaptive Algorithm and its Analysis

Present variable step-size LMS algorithms generally establish a functional relationship between the step-size and the error signal,in order to enhance the convergence and track performance of them.However,they don't take the influence introduced by the input signal into account. So when the input signal changes,the steady state error will increase obviously.This paper basing on some former variable step-size LMS adaptive filtering algorithms,proposes an improved algorithm which introduces the input signal factor and modifies the step-size according to the instantaneous error signal and the input signal.The improved algorithm not only has relative convergence and track performance,but ensures the small steady state error.Simulations show that the improved algorithm performance is better than that of former algorithms.