Real time implementation of multichannel EEG cleaning from muscular artifacts in brain computer interfacing (BCI) paradigm

Abstract

Electroencephalogram (EEG) can re ect not only the activity which is speci c for a cognitive task during an experiment but also the electrical background ac- tivity of the brain. The EEG signals recorded from the scalp surface are usu- ally contaminated by di erent physiological signals which are termed artifacts. Among the artifacts the electrooculography (EOG) makes serious obstacle to many neuroscience experiments including the application for brain-computer interfaces (BCIs). It has noticeable higher energy at lower frequency compared to target EEG signals. This research presents a hybrid wavelet based algorithm to suppress the ocular artifacts from EEG signals. The hybrid wavelet transform (HWT) method is designed by the combination of discrete wavelet decomposition (DWT) and wavelet packet transform (WPT). The artifact suppression is performed by the selection of subbands obtained by HWT. Fractional Gaussian noise (fGn) is used as the reference signal to select the subbands containing the artifacts. The multichannel EEG signal is decomposed by HWT into a nite set of subbands. The energies of the subbands are compared with fGn of the desired subband sig- nals. The EEG signal is reconstructed by the selected subbands consisting of EEG components. The experiments are conducted for both simulated and real EEG signals to study the performance of the proposed algorithm. The results are compared with recently developed artifacts suppression algorithms like stationary subspace analysis (SSA), discrete wavelet transform (DWT) and empirical mode decomposition (EMD). However, the existing EMD method has been successfully used in the eld EEG artifact suppression using a data-adaptive subband ltering approach. But the higher computation burden of EMD processing is the main obstacle to online implementation of brain-computer interfacing (BCI). To resolve such limitation, the proposed HWT with higher computation speed is introduced in this study. In this thesis, BCI experiment is conducted to test the cleaning performance followed by the BCI classi cation with EEG signal. For the motor imagery EEG classi cation, linear discriminant analysis (LDA) is used. The ex- perimental results prove that the classi cation performance increases noticeably with the cleaned EEG data, using the proposed algorithm. It is found that the proposed method performs better than the methods compared in terms of perfor- mance metrics (signal to artifact ratio and mutual information), computational cost and classi cation accuracy. Therefore, this study shows that the proposed HWT based system performs better than other techniques for elimination of blink contamination from EEG signal.