Designing Efficient Blind Source Separation Methods for EEG Motion Artifact Removal Based on Statistical Evaluation

As the electroencephalography (EEG) biomedical signals are affected under the presence of the muscular motion artifacts. Presence of these artifacts leads to error in visual analysis of EEG signal, thus results in wrong diagnosis of human diseases. The variants of blind source separation (BSS) methods are available. This paper aims to design the efficient BSS based method for effectively eradicating the EEG motion artifacts. This is accomplished by evaluating the six different methods, which are combination of independent component analysis (ICA) and canonical correlation analysis (CCA) along with the discrete wavelet transform and stationary wavelet transform methods. Each of above combination methods are applied on the ensemble empirical mode decomposed, Intrinsic Mode Functions for EEG motion artifact suppression. This research paper tests the performance over pure EEG signal and also on the simulated EEG sinusoids to mimic the effect of motion artifacts. The performance of six BSS artifact removal algorithms are evaluated using efficiency matrices such as del signal to noise ratio, lambda (λ), spectral distortion (P_dis) and root mean square error. The execution time is also calculated to evaluate the computation efficiency of the algorithms. The results suggest that CCA algorithm outperforms over ICA in the case of the high noisy condition of EEG signal.