一种不用先验知识的单路混合信号的盲源分离新方法  

A Single-Channel Mixed Signal BSS New Method Without Using the Prior Knowledge

作　　者：成谢锋[1] 马勇[2] 张学军[1] 刘琚[3] 张少白[1] 

机构地区：[1]南京邮电大学电子科学与工程学院,江苏南京210001 [2]南京理工大学计算机学院,江苏南京210094 [3]山东大学信息科学与工程学院,山东济南250001

出　　处：《电子学报》2011年第10期2317-2321,2330,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61073115;No.61076073;No.60806027);南京邮电大学基金(No.NY207139;No.NY210036;No.NY210068)

摘　　要：基于单路混合信号的分层技术,提出了一种单路混合信号的盲源分离新方法,它的特点是不需要利用先验知识.如何做单路原始信号的分层工作?它应该被分几层?通过对分层技术的深入讨论和加入一个参考信号,上述问题找到了很好的答案.并且文中给出了一种改进型圆周卷积数据等长度分层的方法和获取独立子波函数的技术;然后通过结合独立子波函数进入原始信号,使原始信号由一维向量转化成为多维向量,从而实现其盲分离;最后的分离实验,验证了本方法的有效性和可行性.Based on the single-channel mixed signal＇s separation technology and the layered principles,a novel blind source separation（BSS） algorithm of single mixed signal is proposed.its characteristic does not need to use the prior knowledge.How to make the layered work of this original signal？ It can be separated into many layers？ Based on single-channel mixed-signal layered technique and joins a reference signal to this original signal,the above question has found the very good answer.And the equal length layered technique of using improved cyclic convolution decomposition,and the gain independent sub-band functions technology are studied too.Using these independent sub-functions,it can be expanded become a multi-dimensional signal that the original signal,and realized its blind separation.Finally through the separation experiments confirmed this method validity and the feasibility.

关 键 词：盲分离 独立子波函数 单路信号 分层技术 改进型圆周卷积 

分 类 号：TN911.7[电子电信—通信与信息系统]