动态独立分量分析算法及其应用研究

Study on dynamic independent component analysis and its application

对基于非高斯性极大ICA方法的动态实现及其应用进行了研究。介绍了基于峭度极大ICA算法原理,然后将其改造成一种能进行动态独立分量提取的ICA算法。对动态ICA算法的收敛性和盲源分离效果进行了分析研究,并将该算法应用于脑电信号的消噪。结果表明,该算法具有较好的收敛性能和盲源分离效果。

In this paper, the dynamic implementation of independent component analysis （ICA） by maximization of nongaussianity and its application is studied. By introducing the online kurtosis estimation, the dynamic form of ICA algorithm using kurtosis is proposed. In our experiments, the separation performance of dynamic ICA algorithm is compared with those of its batch algorithm. Experiment results show that the dynamic ICA algorithm proposed on this paper has good performance in convergence and source separation.