Robust independent component analysis for blind source separation and extraction with application in electrocardiography

The problems of signal separation and signal extraction arise in a wide variety of applications in biomedical engineering and other areas. Under the source statistical independence assumption, these problems can be solved by independent component analysis (ICA) methods. A simple ICA technique, referred to as RobustICA, has recently been proposed that shows interesting features such as very fast convergence, local-extrema escaping capabilities and the possibility of avoiding prewhitening. The present contribution explains how RobustICA can easily be modified to target particular sources according to their impulsive character as measured by the kurtosis sign. This new feature makes it possible to extract the sources of interest only, or a subspace thereof, with the subsequent reduction in computational complexity and error accumulation. The performance of this modification is illustrated on signal recordings issued from electrocardiography.