自适应步长EASI算法研究及改进  

Research and Improvement of Blind Source Separation Based on EASI Algorithm

作　　者：余华[1] 吴文全[1] 刘忠[1] 

机构地区：[1]海军工程大学电子工程学院,武汉430033

出　　处：《舰船电子工程》2009年第10期118-121,共4页Ship Electronic Engineering

基　　金：国家"八六三"项目"基于多智能体的传感器网络协同目标跟踪技术研究"(编号:2007ADA299)资助

摘　　要：EASI算法(一种在线LMS)是很重要的盲信号分离(BSS)算法,总结了EASI算法应用于盲信号分离的优点与不足,并提出了利用模拟退火策略有效地结合两种步长的优点,在初期快速下降,在后期精确收敛。结果表明,在该实验中使用模拟退火策略来调整学习率简单有效。EASI agorithm is very important in the Blind Signal Seperation（BSS） area, which is a kind of on-line LMS agorithrn. EASI agorithm is further studied using cross talking error as an evaluation parameter. Smulation result and performance comparison of EASI agorithm is given. Validity of EASI agorithm is proved and the advantages and shortcomings of it is discussed, then a new method which using genetic algorithm to adjust step-length is presented. The method integrates the advantages of large step-length and short step-length. Simulation result of the method shows it is simple and effective.

关 键 词：盲信号分离 EASI算法 串音误差(ETC) 模拟退火策略 

分 类 号：TN911.7[电子电信—通信与信息系统]