基于QR分解的最大负熵盲分离算法

Blind separation algorithm of maximal negentropy based on QR decomposition

提出了一种新的实时线性混叠信号盲分离算法。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,并结合源信号相互独立时负熵最大的特点而导出。该方法避免了目前许多学习算法中矩阵逆的计算,从而大大地减少了分离的计算量。理论分析与仿真结果表明该算法不仅具有很好的分离效果,而且也减少了分离时间,其效果均优于Andrzej(1996)和Pham(1999)的相应结果。

A real-time linear mixture signals blind separation algorithm is proposed in this paper.,Firstly,the mixture signals are whiten so that the mixture matrix transform into an orthogonal matrix,according to QR decomposition theory, considering the characteristic that negentropy is maximization when the original signals are independent,then this algorithm is deduced, which can avoid computing inverse matrix that most algorithms must compute recently, so the computational complexity is decreased greatly. Theory analysis and simulation results show that the separation time is reduced and the separation effect is very good. Compared with Andrzej(1996) and Pham(1999), the effect of this algorithm is best.