Online EMG artifacts removal from EEG based on blind source separation

Electromyography (EMG) artifacts are the main and serious contaminated sources to the electroencephalogram (EEG) signals. In this paper, a fully automated EMG removal technique based on canonical correlation analysis (CCA) method is presented. CCA method was proved more suitable to reconstruct the EMG-free EEG data than independent component analysis (ICA) methods in the study. Specially, a number of contaminated and clean EEG data were analyzed in order to decide a reasonable correlation threshold, by which this method can remove successfully not only the light EMG artifacts but also heavy EMG artifacts from the EEG data in real-time application with the little distortion of not only the underlying ictal activity signal but the EOG artifacts.