ONLINE REMOVAL OF EYE BLINK ARTIFACT FROM SCALP EEG USING CANONICAL CORRELATION ANALYSIS BASED METHOD

Abstract

Eye blink artifact, the main contamination in electroencephalography (EEG), brings serious problems for the analysis of EEG data. In this paper, an online method for eye blink artifact removal is presented. Canonical correlation analysis (CCA) is used to decompose the recorded signals containing several-channel EEG and one-channel vertical electrooculography (EOG). The identification of the artifactual component is fully automatically implemented based on evaluating the similarity between the reference EOG and decomposed CCA components. This method was compared with an independent component analysis based technique on a synthetic data set and achieved comparable performance for removing eye blink artifact. Moreover, the CCA based method is less time-consuming. The proposed method was finally implemented with Labview for removing eye blink artifact in online test. The online experiment results show that the proposed method could fulfill the identification and suppression of eye blink artifact from contaminated EEG in real-time.