The Recognition and Removal of Ocular Artifact

Abstract

Eye movements and eye blinks are inevitable in Electroencephalography (EEG) collection experiment and result in ocular artifacts (OA). OA has larger amplitude and is one of main noises of EEG signal. The OA removal effect is important to its interpretation and applications. Based on discrete wavelet transform (DWT) and canonical correlation analysis (CCA), a new and generally applicable method, denoted as DWT-CCA, was proposed to recognize and remove a wide variety of OA from EEG records. DWT was applied to the raw EEG and electrooculogram (EOG) signals to acquire the multi-scale wavelet coefficient and their auto correlation were removed by CCA; Then, the ocular artifacts were recognized by using correlation coefficient and eliminated by setting the corresponding canonical wavelet coefficient vectors to zeros. Furthermore, the inverse algorithms of CCA and DWT were applied successively to obtain the corrected EEG signal. Based on various kinds of OA data produced by horizontal ocular movement, vertical ocular movement, ocular rolling and ocular blinking, the digital experiment research was conducted to compare DWT-CCA with the referenced methods. The experiment results show that DWT-CCA has perfect performance indexes, such as root mean square error, signal noise rate and time consumption, and the statistical tests illustrate significantly the
adaptability for multi-subjects.