自适应噪声对消的归一化LMS算法  

Normalized LMS Algorithm and Its Application in Adaptive Noise Cancellation

作　　者：周龙龙 胡启国[1] ZHOU Longlong;HU Qiguo(School of Mechanical and Electrical Engineering and Vehicle Engineering,Chongqing Jiaotong University,Chongqing 400074,Chin)

机构地区：[1]重庆交通大学机电与车辆工程学院,重庆400074

出　　处：《电声技术》2018年第5期74-77,共4页Audio Engineering

基　　金：国家自然科学基金资助项目(51375519)~~

摘　　要：为了克服自适应滤波中固定步长LMS算法存在收敛速度与稳态误差的矛盾,本文通过MATLAB仿真不同步长因子下LMS算法的学习曲线,分析了LMS算法在收敛过程中存在的矛盾,并运用归一化LMS(NLMS)算法来改善上述矛盾。NLMS算法是通过输入变量改变步长因子从而改变算法的收敛特性。本文对NLMS与LMS算法的误差曲线仿真并进行稳态误差效果比较,结果显示NLMS算法的稳态误差精确度明显提高,收敛速度加快。通过将LMS算法与NLMS算法应用于自适应噪声对消中,得到NLMS算法具有收敛速度更快同时稳态误差更小的特性,该算法能够快速对干扰信号作出反应,使除噪效果更好。In order to overcome the contradiction between convergence speed and steady - state error of the fixed - step LMS algorithm in adaptive filtering,it analyzed the contradiction of LMS algorithm in the convergence process by simulating the learning curve of LMS algorithm in MATLAB. Then it used the normalized LMS （NLMS） algorithm to reduce the above contradictions. NLMS algorithm is an algorithm to change the convergence property of the algorithm by changing the step factor by input variable. In this paper, the error curves of NLMS and LMS algorithm were simulated and the curves＇steady - state errors were compared. The results shown that the steady - state error accuracy of NLMS algorithm was improved obviously and the convergence speed was accelerated. By applying LMS algorithm and NLMS algorithm to adaptive noise cancellation, it concluded that NLMS algorithm had faster convergence speed and smaller steady - state error. So it shown that NLMS algorithm can respond to the interference signal quickly, and make the noise reduction better.

关 键 词：归一化LMS算法 稳态误差 收敛速度 步长因子 权矢量 噪声对消 

分 类 号：TN911.4[电子电信—通信与信息系统]