Blind separation of biosignals by a novel ICA algorithm based on information theory

The biosignals measured by multi-sensors are always the mixtures of several independent sources. Therefore, it is necessary to separate them from each other for clinical diagnosis. According the assumption of statistical independence, the authors propose a novel ICA algorithm based on a mutual information minimization criterion using Edgeworth expansion. The algorithm can result in independent component outputs, which are the recoveries of source signals. Simulation results with EGG signals show the validity of the proposed algorithm.