Multi-dimensional delay-correlation MUSIC: a new method to extract multi-sources of EEGs

The reported studies on EEG inverse by multiple signal classification (MUSIC) show the classical MUSIC algorithm suffers from two shortcomings: be sensitive to a color noise and fail in identifying synchronously active sources. Recent studies reveal that the MUSIC-based algorithm may be improved in depressing spatial coherent noise if the classical zero delay correlation matrix is replaced by a non-zero delay-correlation matrix, or by a high-order cumulant matrix, or by incorporating known noise covariance matrix in the zero-delay correlation matrix. And the MUSIC algorithm can be extended to identify synchronous actives through a recursive strategy. In this work, an iterative, multi-dimensional and delay-correlation MUSIC is proposed, where the color noise is depressed by the non-zero delay correlation and the synchronous active sources are identified by the iterative multi-dimensional MUSIC search. Simulation and VEP data tests show a good reconstruction is obtained.