基于旋转变换的最小互信息量盲分离算法

Blind Separation Algorithm of Minimal Mutual Information Based on Rotating Transform

一种新的实时线性混叠信号盲分离算法在本文提出 .该算法先采取白化混叠信号将混叠矩阵转换为正交矩阵 ,然后基于QR分解理论 ,将混叠信号进行一系列初等旋转变换 ,并结合源信号相互独立时互信息量最小的特点 ,导出了一种新的自适应盲分离算法 .该方法回避了目前基于信息理论方法中 (如Torkkola 1996 ;Pham 1999;Lee 2 0 0 0以及谭 2 0 0 0等 )对“ln|detw|”的复杂计算 .我们不仅给出了详细的理论证明 ,而且也进行了仿真试验 ,理论分析与仿真结果表明该算法减少了分离时间 ,并具有很好的分离效果 .

A new kind of blind separation algorithm for real time linear mixture signals is presented in this paper.At first,the mixture matrix is changed into orthogonal matrix by using whitened mixture signals.Then based on QR decomposing theory,a series of primary rotating transforms are performed for mixture signals.It is also considered that the mutual information is the least if source signals are independent and a new kind of adaptive blind separation algorithm is educed.This method avoids the complex calculation of |det w | with the present information theory method (e.g.Torkkola 1996,Pham 1999,Lee 2000,as well as Tang 2000).In the paper we not only give the theoretic proof in detail,but also carry out the simulation experiments.Both of theoretic analysis and simulation results show that this algorithm saves the separation time and has a good separation effect.