Estimating evoked dipole responses by MEG/EEG for unknown noise covariance

We develop a method for estimating evoked responses by magnetoencephalography (MEG) and electroencephalography (EEG) under the assumption that the noise is spatially correlated with unknown covariance. The source is modeled as a collection of current dipoles and the head as a spherical conductor. We approximate the time evolution of the evoked response by a series of functions. We derive a goodness-of-fit procedure for this model. We also test if the covariance matrix of the noise only equals the covariance matrix of the noise when the evoked response is present. A numerical example demonstrating the performance of the proposed method is presented.