Investigation of two methods based on Empirical Mode Decomposition for removal of real and contaminated blink from EEG

Electroencephalograph (EEG) is the measurement of the electrical activity of the brain, obtained by recording from electrodes placed on the cortex. It is often affected with various artifacts that have biological and external sources. Eye electrical activity might be considered as the most common artifact, which increase the difficulty in analyzing the EEG and to obtaining clinical information. In this paper, two different methods based on Empirical Mode Decomposition (EMD), for elimination of eye blinks in EEG are compared. First method is based on EMD and entropy at which the noisy signal is decomposed into several intrinsic mode functions (IMF) by EMD and entropy is used as the index of stopping IMF extraction, when the process reach the artifact components. The second method uses EMD for extract artifact reference form the noisy signal and employs it as the artifact input in adaptive filtering. The performance index of this experiment is the correlation coefficient between bands of pure EEG and filtered EEG, which indicates both methods worked almost equal in beta, alpha and theta band however, the method based on combination of EMD and adaptive filtering outperforms the EMD and entropy in delta band.