Filtering of Muscle Artifact from the Electroencephalogram

Abstract

When recorded by surface electrodes, the electroencephalogram (EEG) may contain unwanted signals due to depolarization of scalp muscles and various electrochemical effects at the surface-metal junction. The former artifacts, in particular, are difficult to remove by linear filtering. This design study indicates that a state-of-the-art non-linear filter which includes a matched-filter detector with likelihood-ratio decision logic can give significant performance improvement over a third-order linear low-pass filter typical of existing equipment. The same approach can be applied to related electrophysiological filtering problems.