一种可用于低信噪比环境的变步长LMS算法  

A New Variable Step Size LMS Adaptive Filtering Algorithm

作　　者：周映[1] ZHOU Ying(No.10 Research Institute,China Electronics Technology Group Corporation,Chengdu 610036)

机构地区：[1]中国电子科技集团公司第十研究所,成都610036

出　　处：《舰船电子工程》2024年第9期70-73,148,共5页Ship Electronic Engineering

摘　　要：论文研究了自适应最小均方误差(Least Mean Squares,LMS)滤波算法的步长选取问题。在分析现有算法的基础上,通过构造步长与误差信号之间的非线性函数,提出一种新的变步长LMS算法。新算法采用误差信号的自相关估计值控制步长,而不是直接利用瞬时误差控制步长,避免了噪声干扰,降低了稳态失调,可工作于低信噪比环境。同时新算法步长控制无记忆效应,提高了收敛速度。仿真表明,新算法的稳态失调和收敛速度均优于现有变步长LMS算法。The problem of step size selection for adaptive least mean squares(LMS)filtering algorithm is studied.Current algorithms of variable step size LMS adaptive filtering is discussed.By constructing a nonlinear function between the step factor and the error signal,a new algorithm is proposed.New algorithm which makes use of the auto-correlation function of the error signal to control the new variable step size avoids using the instantaneous error energy as a measure to sense the state of adaptation process.As a result,it is robust and can work in low SNR.The level of misadjustment is low.Due to its virtue of no memory in step control,new algorithm provides fast speed of convergence.The results of computer simulation show that new algorithm outperforms all other algorithms in both convergence speed and misadjustment.

关 键 词：自适应滤波 变步长 LMS算法 收敛速度 稳态失调 误差信号 自相关估计 瞬时误差 

分 类 号：TN91[电子电信—通信与信息系统]