Blink artifact elimination in electroencephalographic records based on discrete cosine transform domain modeling

A method for eliminating blink artifacts that contaminate electroencephalographic (EEG) records is introduced. The proposed method enables a direct estimation of blink artifacts without the use of the simultaneously recorded electroculargraphic (EOG) signal. The estimation of blink artifacts in the EEG was achieved by modeling in the discrete cosine transform (DCT) domain. The DCT coefficients of the blink artifact were sufficiently represented by a system transfer function with a low model order of 4. Two cases were investigated: one was blink artifact elimination for the background EEG and the other was for the somatosensory evoked potentials. The effectiveness of the proposed method was demonstrated in an experimental study based on actual EEG data of 11 healthy subjects. The simplicity and effectiveness of the proposed method suggest that blink artifact elimination using DCT domain modeling can possibly be applied in various areas of neuroscience for improving EEG quality.