A new approach for an effective eye movement artifact elimination from EEG signal

EEG signals collected non-invasively from the cerebral cortex are often contaminated with signals from eye movement [EOG], which considerably degrades the reliability of extracting proper information from the actual signals. This study attempts a two-step algorithm to obtain an EOG artifact free EEG signal with best performance index. Two types of decomposing methods namely Discrete Wavelet Transform (DWT) and Empirical Mode Decomposition (EMD) were used initially to decompose the contaminated signal. The resultant signal further undergoes two processes each; first by Canonical Correlation analysis (CCA) technique and the second one by reduced frequency components using threshold level to identify the unwanted frequency components (TH.-FFT). The output signal from each of these two processes are reconstructed to obtain a clean signal. Comparison between the reconstructed cleaned signals and the contaminated EEG signal reveal that the reconstructed signal is cleaner when (CCA) technique was used as compared with that of (TH.-FFT). Thus the performance indexes of this experiment for the two methods, measured with RMSD and SDR, indicated that combined DWT, EMD and CCA method outperforms the combined DWT, EMD and TH.-FFT method in elimination of eye movement contamination from EEG signal. The present algorithm gave better performance results compared to other existing methods and hence would be better suited for various offline analyses involving the low frequency bands of EEG signals.