Robust batch algorithm for sequential blind extraction of noisy biomedical signals

In many applications, especially in biomedical signal processing (like EEG/MEG time series), source signals are noisy and some have kurtosis close to zero. Most known algorithms for blind signal extraction fail to separate all desired sources if the kurtosis is very low or equals zero. In this paper we propose a new simple second order statistic batch algorithm which is able to efficiently extract various temporally correlated sources even if they have very small or even zero kurtosis, for example colored Gaussian sources. Computer simulation examples illustrate validity and performance of the proposed approach for noisy biomedical signals.