一种改进的变步长LMS自适应滤波算法  

A MODIFIED VARIABLE STEP SIZE LMS ADAPTIVE FILTERING ALGORITHM

作　　者：石嘉豪[1] 罗雅愉[1] 刘威杨[1] 梅剑寒[1] 王为凯[1] 

机构地区：[1]华南理工大学电子与信息学院广东广州510641

出　　处：《计算机应用与软件》2013年第9期183-186,共4页Computer Applications and Software

基　　金：广东省创新实验项目(1056111035)

摘　　要：针对最小均方算法收敛过程中收敛速度与稳态误差的矛盾,提出一种基于反正切函数的归一化最小均方算法。该算法利用反正切函数和误差自相关的时间估计建立了步长与误差之间的非线性关系,抑制环境中的非相关噪声,同时引入归一化信号功率扩大输入信号的取值。仿真结果表明,该算法具有较快的收敛速度、较低的稳态误差,同时具备较好的系统时变跟踪能力。Aiming at the contradiction between convergence rate and static error in convergence process of the least mean square （LMS） algorithm, a normalised LMS algorithm based on arctangent function is proposed. In this algorithm, arctangent function and time estimation of the error autocorrelation are utilised to establish a non-linear relation between the step factor and the error to suppress the non-correlated noise in the environment; meanwhile, the normalised signal power is introduced to expand the value range of input signal. Simulation results confirm that the proposed algorithm has faster convergence rate, lower static error and has better capability of system time-varying tracking.

关 键 词：自适应滤波 变步长 最小均方 反正切函数 

分 类 号：TP302.7[自动化与计算机技术—计算机系统结构]