一种新的变步长LMS算法  

Novel variable step size LMS algorithm

作　　者：张永华[1] 马社祥[1] 代红维[1] 

机构地区：[1]天津理工大学计算机与通信工程学院,天津300384

出　　处：《天津理工大学学报》2010年第1期62-64,共3页Journal of Tianjin University of Technology

摘　　要：对已有的一些变步长LMS自适应滤波算法进行了分析,在此基础上提出一种改进的变步长LMS算法.该算法建立了步长因子与误差信号和权系数变化之间的非线性函数关系,从而使权向量异步更新达到最佳.仿真结果表明,该算法具有更快的收敛速度,更小的稳态误差及更平稳的收敛过程.Some variable step size LMS algorithms in literature are analyzed, based on which an improved one is presented. This algorithm builds a nonlinear relationship between the step size, the error signal and the weighting coefficients, and then the asynchronous update of the weight vector becomes optimal. Simulation results show that this algorithm has higher speed of convergence, lower steady state error and more stable convergence process.

关 键 词：LMS算法 变步长 收敛速度 稳态失调 

分 类 号：TN911.23[电子电信—通信与信息系统]