Improved techniques for independent component analysis
Yongjian Zhao, Xiaoming Kong, Haining Jiang, Meixia Qu · 2014
The traditional independent component analysis (ICA) approach aims to separate all underlying independent components (original sources) from their mixtures simultaneously. Through introducing specific prior information about a source signal, an improved ICA algorithm is derived in this paper. One can extract a desired source signal while its normalized kurtosis range is known in advance. The utility of the proposed algorithm is demonstrated by computer simulations on biomedical signals.