quotation:		[Copy]
		[Copy]

This Paper:Browse 1548 Download 1899
基于稀疏系统辨识的改进的零吸引LMS算法
辛龙坤,孟金,易胜宏,刘挺
0 Fontlarge +\|Default\|Small
(重庆邮电大学通信与信息工程学院,重庆 400065;重庆邮电大学光电工程学院,重庆 400065)

摘要:

针对当前的稀疏最小均方(Least Mean Square,LMS)算法普遍存在的收敛速度和稳态均方差(Mean Square Deviation,MSD)不能同时达到一个较好状态的问题，提出了一种改进的零吸引最小均方(Improving Zero-attracting LMS,IZA-LMS)算法。在滤波器估计较大或较小的冲激响应时，IZA-LMS算法的零吸引函数分别采用重新加权的零吸引LMS(Reweighting ZA-LMS，RZA-LMS)算法的零吸引函数和改进的l0-norm惩罚函数，使算法同时满足较快的收敛速度和较低的MSD值。理论分析和仿真证明，IZA-LMS算法比目前的诸多稀疏LMS算法的收敛速度更快且稳态MSD更低。

关键词: 自适应滤波稀疏系统辨识最小均方算法零吸引稳态均方差

DOI：

基金项目:

An improved zero-attractive LMS algorithm based on sparse system identification

XIN Longkun,MENG Jin,YI Shenghong,LIU Ting

(School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;School of Optoelectronics Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

Abstract:

To solve the problem that the current sparse least mean square (LMS) algorithm generally can’t reach the better state of convergence speed and steady-state mean square deviation (MSD) at the same time,an improved zero-attracting LMS(IZA-LMS) algorithm is proposed.When the filter estimates large or small impulse responses,the ZA function of the IZA-LMS algorithm both adopts the ZA of the reweighted zero-attraction LMS (Reweighting ZA-LMS,RZA-LMS) algorithm and the improved l0-norm penalty function respectively so that it meets both faster convergence speed and lower MSD values.Theoretical analysis and simulation result prove that the IZA-LMS algorithm converges faster and has lower steady-state MSD than many current sparse LMS algorithms.

Key words: adaptive filtering sparse system identification least mean square zero-attracting mean square deviation