Extraction and separation of eyes movements and the muscular tonus from a restricted number of electrodes using the independent component analysis

Up to now the diagnosis of sleep apnea syndrome (SAS) is based on the analysis of a polysomnogram. This latter however is costly and uncomfortable for the patient. Therefore, we are working on the elaboration of a small data acquisition system. This communication focuses on a method for recovering from a reduced number of sensors informative signals to establish a diagnostic. The proposed method makes use of the independent component analysis (ICA). Several methods of ICA were applied on electroencephalogram (EEG) and magnetoencephalogram (MEG) signals for artifacts removal. On the contrary to previous works, ICA method in this paper is shown to extract separately the EEG, the electroocculogram (EOG) and the muscular activity (EMG) from a very restricted number of electrodes. Moreover, since we are confronted with an underdetermined problem in ICA (less observations than sources), increasing the number of mixing observations by filtering the signals issued from the restricted channels is demonstrated to be an efficient solution.