首页 | 本学科首页   官方微博 | 高级检索  
     

步长随输入信噪比变化的LMS算法
引用本文:焦君圣,王昌明,杜栓平,王忠康.步长随输入信噪比变化的LMS算法[J].数据采集与处理,2005,20(2):139-143.
作者姓名:焦君圣  王昌明  杜栓平  王忠康
作者单位:南京理工大学机械工程学院精密仪器系,南京,210094;杭州应用声学研究所海洋声学国家级重点实验室,杭州,310012
摘    要:介绍了归一化最小均方误差(LMS)算法,分析了它在平稳环境噪声下步长不能自动调整的弊端。然后提出了步长随输入信噪比变化的LMS算法,分析了步长调整规律。根据该算法在噪声抵消的起始阶段收敛性能差并可能出现发散的现象,对该算法进行了修正,即在起始阶段采用归一化LMS算法,之后利用步长随信噪比变化的LMS算法。通过对语音信号的消噪仿真,表明修正的步长随输入信噪比变化的LMS算法收敛性能好,输出信号的信噪比有了较大的提高。

关 键 词:最小均方误差  归一化LMS  输入信噪比  自适应噪声抵消器
文章编号:1004-9037(2005)02-0139-05
修稿时间:2004年7月10日

LMS Algorithm for Variable Step-Size Adjusting with Input Signal-to-Noise Ratio
JIAO Jun-sheng,WANG Chang-ming,DU Shuan-ping,WANG Zhong-kang.LMS Algorithm for Variable Step-Size Adjusting with Input Signal-to-Noise Ratio[J].Journal of Data Acquisition & Processing,2005,20(2):139-143.
Authors:JIAO Jun-sheng  WANG Chang-ming  DU Shuan-ping  WANG Zhong-kang
Affiliation:JIAO Jun-sheng1,WANG Chang-ming1,DU Shuan-ping2,WANG Zhong-kang2
Abstract:The normalized LMS (NLMS) is presented. The step-size can not be automatically adjusted according to the input SNR in the stationary noise environment. Then the LMS algorithm for the variable step-size is proposed, which is adjusted according to the input SNR. Finally, the step-size adjusting rules are analyzed. During the initial phase of noise cancellation, its step-size can not satisfy with the requirement. The step-size should decrease with the iterations to achieve the rapid convergence. It may even lead to the divergence. So at the initial stage, NLMS is adopted and then the algorithm is proposed. Simulations show that adopting this combined algorithm in the noise cancellation of the voice signal results in significant output SNR to be improved.
Keywords:LMS  normalized LMS  input SNR  adaptive noise canceller
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号