一种盲信号分离的信息理论方法

AN INFORMATION THEORY METHOD FOR BLIND SEPARATION OF SIGNALS

给出了一种新的基于信息理论的盲信号分离的判据，该判据基于统计独立的假设，同时利用了最大信息传输和输出互信息最小化．将该判据用于独立分量分析，得到源信号的盲分离．仿真实验结果表明，同ｉｎｆｏｍａｘ方法相比，该方法可分离范围更广的信号．

A new information theory based criterion is proposed for blind separation of signals in this paper,where the maximization of information transfer and minimization of mutual information of the output are applied.A separation matrix is derived while the criterion used to ICA.The computer simulation shows that the proposed method can separate some mixtures which infomax method would not.