Abstract

In the last decade, large arrays of sensors for magnetoencephalography (MEG) (and electroencephalography (EEG)) have become more commonplace, allowing new opportunities for the application of beamforming techniques to the joint problems of signal estimation and noise reduction. We introduce a new approach to noise cancellation, the generalized sidelobe canceller (GSC), itself an alternative to the linearly constrained minimum variance (LCMV) algorithm. The GSC framework naturally fits within the other noise reduction techniques that employ real or virtual reference arrays. Using expository human subject data with strong environmental and biological artifacts, we demonstrate a straightforward sequence of steps for practical noise filtering, applicable to any large array sensor design.