后非线性混合盲信号分离的一种新算法  

A New Algorithmfor Blind Separation of Post-Nonlinear Mixture

作　　者：柳艳红[1] 魏学业[1] 吉松坡 

机构地区：[1]北京交通大学电子信息工程学院,中国电信新疆电信有限公司,北京10044,新疆乌鲁木齐830026

出　　处：《电子学报》2008年第7期1401-1404,共4页Acta Electronica Sinica

摘　　要：本文提出了一种新的后非线性混合盲信号分离算法.现存算法大多需要额外的附加源信号信息,才能实现信号的分离,使盲分离变成了半盲分离.鉴于此,本文提出了一种不需要任何附加信息的全盲分离算法.首先,通过微分变换将后非线性混合模型变换成形式如同线性瞬时混合模型的形式,并论证了源信号的微分形式保留了源信号的统计特征.这样,就使非线性问题得到大大简化.其次,利用信号的相关特性建立目标函数及递推方式,用LMS算法使目标函数达到最小值,从而实现了盲信号分离的目的.最后,通过计算机仿真试验验证了本文算法的可行性和有效性.与现存算法相比,本文算法计算量小,收敛速度快,实时性好,实现了全盲分离.A new algorithm for blind separation of post-nonlinear mixture is proposed in this paper. The extra information on source signals is required and blind separation becomes half-blind separation for the past-nonlinear mixture in the most of the existing algorithms. So whole-blind separation algorithm without any extra information is proposed. Fisstly, the model of post-nonlinear mixture is transformed into the model like the instantaneous linear mixture by the differential transformation, which makes the nonlinear problem simplified largely. Then it is goofed that the differential of source signals conserves the same statistical characters as the source signals. Secondly, the correlation characters of signals are utilized to construct separation criterion function and iterative equation, then LMS algorithm is utilized to make criterion function minimized for blind source separation. Lastly, the numerical computer simulations are performed to illustrate the validity and feasibility of our algorithm. Compared with the existing algorithms, our algorithm possesses some good excellences like low computational complexity, rapid convergence, well real-time processing and whole-blind source separation.

关 键 词：盲信号分离 后非线性 去相关 LMS算法 实时处理 

分 类 号：TN911[电子电信—通信与信息系统]