EMG Artifacts Removal from Multi-Channel EEG Signals using Multi-Channel Singular Spectrum Analysis

Abstract

The Electroencephalogram (EEG) is the brain signals which are most normally debased by Electromyogram (EMG) antiquities. The presence of these EMG antiquities covers the necessary information in an EEG signal. In this paper, we have proposed another strategy named as Multi-channel Singular Spectrum Analysis (MSSA) in light of Singular Value Decomposition (SVD) to expel muscle or EMG antiquities from multi-channel EEG signals. At first, the orthogonal eigenvectors of multi-channel data are estimated by performing SVD which are acquired from the covariance matrix. Since the frequency variations of eigenvectors related to EEG signal are quite low when compared to the EMG signal, so we fix some peak frequency threshold to find out the frequencies related to EEG signal, then the frequencies related to EMG signals are suppressed and the artifact free Multi-channel EEG signal is extracted. Finally, our proposed technique is applied on a noisy sinusoidal signals to test the performance of the proposed method and then it is applied on synthetic EEG signals mixed with the EMG artifacts. Simulation results are then compared with Canonical Correlation Analysis (CCA) to show that the proposed method eliminates EMG antiquities more adequately without amending the required data.