Automatic removal of EEG artifacts using ICA and Lifting Wavelet Transform

Abstract

EEG artifacts significantly affect the accuracy of feature extraction and data classification of Brain-computer interface (BCI) systems. The EEG artifacts derived from ocular and muscular activities are inevitable and unpredictable due to subject's physical conditions. Consequently, the removal of these artifacts is a crucial function for BCI applications to make the system more robust. One of the most prominent techniques employed to remove the EEG artifacts is Independent Component Analysis (ICA). This technique separates EEG signals into Independent Components (ICs) and then discriminates EEG artifacts from neurally generated brain signals. However, the source separation of ICA algorithm is imperfect. Frequently, the IC identified to be an artifact includes brain wave activities useful for data classification. The proposed method will elaborate on the IC with Lifting Wavelet Transform (LWT) to extract the useful neural signals from the artifact component. Experimental results prove the performance and accuracy of the proposed removal algorithm of light and strong eye-blink artifacts. This removal technique implemented in NECTEC's Neurofeedback System for Attention Training was tested in pre-trial sessions with 10 healthy subjects and 5 MCI patients at Chulalongkorn Hospital, Bangkok.