The presence of electrooculographic (EOG) artifacts in the electroencephalographic (EEG) signal is a major problem in the study of brain potentials. A variety of algorithms have been proposed to reject these artifacts including methods based on regression and blind source separation (BSS) techniques. None of them has so far been established as the method of choice. In the present study, the performances of five widely used EOG artifact rejection techniques are compared. The compared methodologies include two fully automated regression methods, one based on Least Mean Square (LMS) for its optimization process, and the other on Recursive Least Square (RLS) algorithm, two BSS techniques which use respectively the Extended — Independent Component Analysis (ext — ICA) and the Second Order Blind Identification (SOBI), and finally a time-varying adaptive algorithm based on H ∞ principles (H ∞ — TV). Each algorithm was applied in real EEG data and then their performance quantified in the time domain. The performance of RLS and H ∞ — TV were poor in removing eye — blink artifacts. For the rest of the methods the results supported the use of LMS technique and suggested the need for further research examining the performance of various artifact rejection techniques in both time and frequency domain.
|Title of host publication||4th European Conference of the International Federation for Medical and Biological Engineering|
|Publication status||Published - 2009|