Neuroelectromagnetic imaging of correlated sources using a novel subspace penalized sparse learning

Brain signal source localization from E/MEG has been an active research area. Currently, there exists var- ious approaches such as MUSIC and M-SBL. However, when the unknown sources are highly correlated, conventional algorithms often exhibit spurious reconstructions. To address the problem, we propose a new algorithm that generalizes M-SBL by exploiting the fundamental subspace geometry in the multiple measurement problem (MMV). Results show that the proposed method outperforms the existing methods even with a highly correlated source.