Automatic EOG Artifact Removal in Brain-computer Interface Systems

Abstract

In this study, we propose a system to recognize the finger-lifting electroencephalogram (EEG) data. Combined with independent component analysis (ICA) and feature extraction, fuzzy c-means (FCM) clustering is used to discriminate between left and right finger movement without supervision. ICA is used to eliminate the electrooculography (EOG) artifacts. Wavelet-fractal features are then extracted from wavelet data via fractal dimension. FCM clustering is used for feature discrimination. It is an unsupervised approach suitable for the applications of biomedical signals. After EOG artifact removal, the performance is improved for all subjects.