稀疏偏差补偿最小平均对数算法  

Sparse Bias-compensated Least Mean Logarithmic Square Algorithm

作　　者：王学成[1] 张佳庚 马文涛[3] WANG Xuecheng;ZHANG Jiageng;MA Wentao(School of Information and Engineering,Shaanxi Institute of international Trade & Commerce,Xi'an 712046,China;Network Information Center,Xi'an Jiaotong University,Xi'an 710049,China;School of Automation and Information Engineering,Xi'an University of Technology,Xi'an 710048,China)

机构地区：[1]陕西国际商贸学院信息与工程学院,陕西西安712046 [2]西安交通大学网络信息中心,陕西西安710049 [3]西安理工大学自动化与信息工程学院,陕西西安710048

出　　处：《信息与控制》2019年第4期413-419,共7页Information and Control

基　　金：陕西省自然科学基础研究计划资助项目(2017JM6033);陕西省教育厅科研计划项目资助项目(17JK0550);咸阳市科技成果推广计划资助项目(2015KT-15)

摘　　要：针对最小平均对数(LMLS)算法在输入信号受噪声干扰的环境下进行稀疏系统辨识时存在精度低的问题,提出了一种稀疏偏差补偿LMLS算法.利用无偏准则推导偏差补偿项来修正输入噪声带来的偏差,构建偏差补偿LMLS.借助系统稀疏特性的先验知识,采用互相关熵诱导维度作为稀疏惩罚约束条件,优化偏差补偿LMLS算法.仿真结果表明,所提算法对于含噪输入信号下的稀疏系统参数辨识具有高稳态精度.We propose a sparse bias-compensated least mean logarithmic square (LMLS) algorithm to solve the low precision problem of the LMLS algorithm in sparse system identification with noisy input. We derive the bias compensation term by using the unbiasedness criterion to compensate the bias caused by input noise, and construct a bias-compensated LMLS (BCLMLS). To fully utilize the prior knowledge on the sparse system, we introduce the correntropy-induced metric as a sparse penalty constraint to optimize the proposed BCLMLS algorithm. The simulation results demonstrate that the proposed algorithm has a good steady-state performance in solving sparse system identification problems with noisy inputs.

关 键 词：稀疏系统辨识 偏差补偿 最小平均对数 互相关熵诱导维度(CIM) 含噪输入信号 

分 类 号：TN911[电子电信—通信与信息系统]