一种新的可变步长LMS自适应滤波算法  

A Novel LMS Adaptive Filtering Algorithm with Variable Step Size

作　　者：王敏强[1] 胡贵龙[1] 郑宝玉[1] 

机构地区：[1]南京邮电学院信息工程系,江苏南京210003

出　　处：《南京邮电学院学报（自然科学版）》2003年第4期12-16,共5页Journal of Nanjing University of Posts and Telecommunications

摘　　要：在简单讨论基本LMS,变步长NLMS和LMS/F组合自适应滤波算法的基础上提出一种新的可变步长LMS自适应滤波算法,新算法引入修正系数ρ和遗忘因子λi=exp(-i),并利用ρ和λi来产生新的步长参与迭代。计算机仿真结果表明,与基本LMS算法或变步长NLMS、LMS/F组合算法相比,新算法在保持算法简单这一特点的同时进一步加快了收敛速度,并能够收敛到更小且稳定的均方误差(MSE)。Following a brief discussion on basic LMS, variable step size NLMS and Combined LMS/F algorithms,a novel adaptive filtering algorithm with variable step size is proposed. The new algorithm introduces a real scaling factor ρ and forgetting factor λi=exp(-i) (i=1,2,…,M-1) to produce the new variable step size μ(n) and for iteration. Computer simulations demonstrate that the proposed algorithm accomplishes faster convergence and steady smaller MSE than the LMS, NLMS and Combined LMS/F algorithms with only a small increase of computational complexity.

关 键 词：LMS 可变步长自适应滤波算法 最小均方 遗忘因子 修正系数 

分 类 号：TN911.7[电子电信—通信与信息系统]