Artifacts-matched blind source separation and wavelet transform for multichannel EEG denoising. (September 2015)

• Record Type: Journal Article
• Authors: Rakibul Mowla, Md; Ng, Siew-Cheok; Zilany, Muhammad S.A.; Paramesran, Raveendran
• Is Part Of: Biomedical signal processing and control. Volume 22(2015)
• Page Start: 111
• Page End: 118
• Publication Date: 2015-09
• Subjects: EEG artifact removal -- Second-order blind identification (SOBI) -- Canonical correlation analysis (CCA) -- Empirical mode decomposition (EMD) -- Stationary wavelet transform (SWT); Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2015.06.009
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract : Highlights: Electromyogram (EMG) and Electrooculogram (EOG) artifacts are major problems in EEG signal analysis. Proposed method is capable of removing both EMG and EOG artifacts. The method uses CCA-SWT for removing EMG artifacts and automatically switches to SOBI-SWT for removing EOG artifacts. Ability to remove artifacts and allow the evoked potential to be detected from actual EEG experiment. Abstract: The physiological artifacts such as electromyogram (EMG) and electrooculogram (EOG) remain a major problem in electroencephalogram (EEG) research. A number of techniques are currently in use to remove these artifacts with the hope that the process does not unduly degrade the quality of the obscured EEG. In this paper, a new method has been proposed by combining two techniques: a canonical correlation analysis (CCA) followed by a stationary wavelet transform (SWT) to remove EMG artifacts and a second-order blind identification (SOBI) technique followed by SWT to remove EOG artifacts. The simulation results show that these combinations are more effective than either using the individual techniques alone or using other combinations of techniques. The quality of the artifact removal is evaluated by calculating correlations between processed and unprocessed data, and the practicability of the technique is judged by comparing execution times of the algorithms.