Source Localization of Brain Electrical Activity via Time-frequency Linearly Constrained Minimum Variance Method

The locations of active brain areas can be estimated from the surface recordings. We describe the localization of the sources of the brain electrical activity via spatial filtering. This method incorporates the time-frequency characteristics of the neural sources for its location. The estimation of the location of active brain area is done in non-parametric fashion. The spatial filters are implemented as weighted sum of data recorded at different sites. The weights are chosen to minimize the output of the filter subject to linear constraint. The function of the linear constraint is to pass the brain electrical signal from a specific location, while attenuating the overall variance at the output of the filter. The method exploits space time-frequency covariance matrix of the data. The proposed algorithm is suited to localizing sources of time-varying and nonstationary signals. For this reason, this paper focuses on the class of frequency modulated (FM) signals, e.g., chirp signals.