Blind Source Separation of Biomedical Signals: convolutive versus instantaneous algorithms

In this paper an algorithm for solving blind separation of convolutive mixtures is introduced. The convolutive model is an extension of the instantaneous one and it allows to relax the hypothesis of a linear mixing process. This convolutive independent component analysis is solved out in the frequency domain, where the algorithms developed for the instantaneous model can be employed with minor modifications. After computing the performance of different algorithms in frequency domain, a comparative evaluation of convolutive model and of the instantaneous linear mixing one is reported. The decomposition performance of such approaches is evaluated on simulated dataset of convolutive mixtures of biomedical signals.