基于网络分量分析的盲源分离方法

Blind source separation based on network component analysis

提出了用先验混合矩阵对盲源进行分离的网络分量分析方法(NCA)。该方法在统计独立性假设不成立的条件下,也能实现对源信号的分离。通过计算机仿真与FastICA和JADE算法进行了性能比较分析,证实了在无统计独立性的假设下,NCA具有更理想的盲源分离效果。

A method of Network Component Analysis （NCA） which separated blind sources using a priori information on the mixing matrix was put forward. Therefore blind source separation could be achieved without the assumption of statistical independence. Performance analysis is given compared with FastICA and JADE through computer simulation. The superiority of NCA is validated without the assumption of statistical independence.