基于Pearson系统模型的盲信号分离研究

RESEARCH ON BLIND SOURCE SEPARATION BASED ON PEARSON SYSTEM MODEL

独立分量分析是盲源分离的主流技术。自然梯度算法是其中非常重要的算法之一。介绍一种最大似然框架下的Pear-son系统模型。该方法的优点是无须知晓信号的概率分布,实验结果表明,该算法能有效地分离随机混合的信号,特别对于非对称源有比同类算法更理想的效果。

Independent component analysis is a mainstream technique for blind source separation. Natural gradient algorithm is one of the most important algorithms. This paper introduces a method for blind source separation without any knowledge of their probability distributions. This is achieved under a maximum likelihood framework by Pearson system model. Simulation result shows that the proposed algorithm is able to separate random mixing signals, especially for asymmetric sources.