Decomposing subspaces of EEG channel space into potentials of non-overlapping distributed sources

Abstract

In many situations various methods to analyze EEG data result in subspaces of the sensor space spanned by potentials of a set of sources. We propose a general model free method to decompose such a subspace into contributions from distinct sources. This unique decomposition can be achieved by first finding the respective subspace in source space using a linear inverse method and then finding the linear transformation such that the source distributions are mutually orthogonal and have a minimum overlap. The corresponding algorithm is extremely efficient and is almost never trapped in local minima. The method is illustrated with results for alpha rhythm. © 2009 Springer-Verlag.