New beamformers for waveform reconstruction of the correlated EEG sources

Abstract

This paper presented a method which may reasonably accurately reconstruct time courses of the correlated sources. This method first uses estimates of the source locations to construct a transformation, which removes the correlated interfering sources while retaining the desired sources. Conventional adaptive beamformers is then performed on the transformed data containing only desired sources to obtain accurate source time course estimates. Repeatedly applying the method to each of the correlated sources can accurately reconstruct all the source time courses. Theoretical analysis shows that the proposed method can remove the leakage from the other correlated sources and thus, obtain accurate source time courses estimates. Numerical experiments show that, compared to the conventional adaptive beamformers, the proposed method can reconstruct the source time courses without distortions.