摘要
在系统辨识、回声消除、即时翻译系统中,背景噪声往往呈现出很强的脉冲特性,传统的自适应滤波算法对脉冲噪声的鲁棒性较差,而基于最大熵准则的自适应滤波算法可以有效地提高脉冲噪声环境下的自适应算法辨识性能.本文提出了一种基于最大相关熵准则的簇稀疏鲁棒仿射投影(Cluster-sparse robust affine projection,CS-RAP)算法,它可以用于辨识回声系统、卫星通信系统等簇稀疏系统.我们在基于最大相关熵准则的仿射投影算法的代价函数中引入权向量的混合L2,1范数约束来利用系统的簇稀疏特性,采用基追踪法来推导CS-RAP算法.最后通过各种仿真实验,来验证提出的CS-RAP算法的鲁棒性和有效性.仿真结果表明,在脉冲噪声环境下,CS-RAP算法与其它相关算法相比具有更快的收敛速度和更低的估计偏差.
In many engineering applications,such as system identification,echo cancelation and real-time translation systems,the noise often shows strong impulsive characteristics,while the traditional adaptive filtering algorithm has poor robustness to impulsive noise,and the adaptive filtering algorithm based on the maximum correntropy criterion can effectively improve the identification performance of the adaptive algorithm in impulsive noise environment.In this paper,a cluster-sparse robust affine projection(CS-RAP) algorithm is proposed,which is used to identify cluster-sparse systems such as echo systems and satellite communication systems.To adequately use the cluster-sparse property,we derive the CS-RAP algorithm by introducing the mixed L2,1 norm constraint of weight vector into the cost function of the affine projection algorithm with MCC.Finally,the robustness and effectiveness of the proposed CS-RAP algorithm are verified by various simulation experiments.The simulation results show that the CS-RAP algorithm has faster convergence speed and lower identification bias than other related algorithms in impulsive noise environment.
作者
黄梓桐
阿里甫·库尔班
韩文轩
刘冰慧
HUANG Zitong;Alifu Kuerban;HAN Wenxuan;LIU Binghui(School of Software,Xinjiang University,Urumqi Xinjiang 830046,China)
出处
《新疆大学学报（自然科学版）》
CAS
2020年第2期177-182,共6页
Journal of Xinjiang University(Natural Science Edition)
基金
国家自然科学基金资助项目(No.61562084).
关键词
系统辨识
最大相关熵
仿射投影算法
混合范数
脉冲噪声环境
system identification
maximum correntropy criterion
affine projection algorithm
mixed norm
impulsive environment
