用于稀疏系统辨识的低复杂度集员NLMS算法  

Low Complexity Set-Membership NLMS Algorithms for Sparse System Identification

作　　者：孟金 易胜宏 张红升[1] 甘济章 雷鹏 MENG Jin;YI Shenghong;ZHANG Hongsheng;GAN Jizhang;LEI Peng(College of Optoelectronic Engineering,Chongqing University of Post and Telecommunications,Chongqing 400065,China)

机构地区：[1]重庆邮电大学光电工程学院,重庆400065

出　　处：《信号处理》2022年第2期417-425,共9页Journal of Signal Processing

基　　金：国家重点研发计划重点专项项目(2019YFC1511300);重庆市技术创新与应用发展项目(cstc2019jscx-msxmX0079)。

摘　　要：针对现有的稀疏集员(SM,Set-Membership)自适应滤波算法,普遍存在稳态均方偏差(MSD,Mean Square Deviation)的稳定性较低及运算复杂度较高等问题,提出了一种新颖的稀疏集员NLMS(NLMS,Normalized Least Mean Square)算法。该方案提出一种运算复杂度更低的零吸引(ZA,Zero-attracting)惩罚函数,使算法对稀疏信道的估计更为合理。在多种稀疏信道仿真中,与目前的稀疏集员算法相比,所提算法的稳态MSD更低且稳定性更高。最后,将所提算法应用于声波信号的稀疏系统辨识中,比目前的稀疏集员算法更具有优势。A novel sparse set membership NLMS(Normalized Least Mean Square)algorithm is proposed to address the existing sparse set-membership(SM)adaptive filtering algorithms,which generally has the problems of low stability of steady-state mean square deviation(MSD)and high operational complexity. The scheme proposes a zero-attraction(ZA) penalty function with lower computational complexity,which makes the algorithm more reasonable for estimation sparse channels. In a variety of sparse channel simulations,the proposed algorithm has lower steady-state MSD and higher stability compared to the current sparse setter algorithm. Finally,the proposed algorithm is applied to the sparse system identification of acoustic signals,which is more advantageous than the current sparse setter algorithm.

关 键 词：稀疏信道 零吸引惩罚 集员 系统辨识 

分 类 号：TN912.3[电子电信—通信与信息系统]