盲信源分离自然梯度算法的仿真  

Simulation of Blind Source Separation Algorithm with Natural Gradient

作　　者：杨海涛[1] YANG Hai-Tao(College of Mathematics,Inner Mongolia University for Nationalities,Tongliao 028043,China)

机构地区：[1]内蒙古民族大学数学学院,内蒙古通辽028043

出　　处：《内蒙古民族大学学报（自然科学版）》2018年第2期118-123,共6页Journal of Inner Mongolia Minzu University：Natural Sciences

基　　金：国家自然科学基金资助项目(11561052)

摘　　要：采用典型的自然梯度算法对盲信源分离进行仿真,通过仿真结果对自适应步长的大小对串音误差曲线的收敛速度以及稳定性的影响进行分析.并根据结果通过文献调研对峭度自适应学习率算法以及EASI算法进行了仿真和讨论,并与传统的自然梯度算法进行了比较和分析.The blind source separation algorithm using the typical natural gradient algorithm was simulated,and the influence of the convergence rate and stability of the crosstalk error curve caused by the adaptive step size through the simulation results was analyzed.According to the results,the kurtosis adaptive learning rate algorithm and EASI Algorithm were simulated and discussed.Comparison and analysis of the natural gradient algorithm with the traditionalone is presented.

关 键 词：盲信源分离 自适应步长 峭度 EASI 

分 类 号：TN911.7[电子电信—通信与信息系统]