一种基于解相关的变步长LMS算法  

New variable step-size LMS-Type algorithm and simulation based on Decorrelation method

作　　者：龙宇[1] 王忠[1] 薛晓娜[1] 

机构地区：[1]四川大学电气信息学院通信工程系,成都610065

出　　处：《电子测量技术》2007年第1期52-55,共4页Electronic Measurement Technology

摘　　要：在自适应滤波器的构造算法中,LMS算法作为经典理论其应用范围广泛,如何进一步提高滤波器的性能,满足实时数据处理的需要,是当前的一个研究热点。本文讨论了一类针对传统LMS算法进行改进的变步长自适应算法,分析其性能,给出了一种新的变步长自适应算法。这种算法同时从调节变步长的形式和滤波器权系数向量的更新方式2个方面出发,进一步提高了算法的收敛速率,充分调和了收敛速率与稳态误差之间的矛盾。计算机仿真结果表明,新的基于解相关的变步长LMS算法具有更快的收敛速率和更小的稳态误差,能较好地对信号进行跟踪滤波。A series of step-size algorithm have been proposed to enhance the performance of LMS-Type algorithm. But experiment based on the actual environment indicates that of these algorithm are not good enough with performance when the signal-to-noise is low. This paper presents a new variable step-size LMS-type algorithm based on the decorrelation method,which pays attention to the type of variable step-size and updates direction of the filter coefficient vector. From two of the aspect, this method increases the convergence velocity, and conciliates the conflict between convergence velocity and reliable error. Finally the simulation is giyen and proves the superiority towards other algorithms.

关 键 词：LMS算法 变步长 解相关 失调 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TP273[自动化与计算机技术—计算机科学与技术]