Removal of EOG artifacts from single channel EEG – An efficient model combining overlap segmented ASSA and ANC

Abstract

Electroencephalogram (EEG) signals are mostly interfered by electrooculogram (EOG) artifacts. These artifacts degrade the performance of portable or wearable EEG recording systems. In this work, overlap segmented adaptive singular spectrum analysis (Ov-ASSA) combined with adaptive noise canceler (ANC) technique is presented for removal of EOG artifacts. Depending on the amplitude of the EEG signal, the first one or two reconstructed components of Ov-SSA technique are adaptively grouped and considered as a reference EOG signal for ANC in a single channel EEG recording system. In order to demonstrate the performance of the proposed technique, Matlab simulations are done on both synthetic and real EEG data. The synthetic EEG data is derived using the Markov Process Amplitude (MPA) EEG model. The performance metrics namely, RRMSE (relative root mean square error) and MAE (Mean Absolute Error) of proposed Ov-ASSA-ANC model outperformed the existing techniques. In addition to the removal of EOG artifacts, the proposed Ov-ASSA-ANC technique was also applied for seizure detection and an average accuracy of 98.05% was achieved.