Automatic ocular correction in EEG recordings using maximum likelihood estimation

Abstract

The electrooculogram (EOG) artifact is one of the main contaminators of electroencephalographic recording (EEG). EOG can make serious problems in results and interpretations of EEG processing. Rejecting contaminated EEG segments result in an unacceptable data loss. Many methods were proposed to correct EOG artifact mainly based on regression and blind source separation (BSS). In this study, we proposed an automatic correction method based on maximum likelihood estimation. The proposed method was applied to our simulated data (real artifact free EEG plus controlled EOG) and results show that this method gives superior performance to Schlögl and SOBI methods.