Research on removing ocular artifact automatically from EEG signals

Electroencephalography (EEG) is easily affected by ocular artifact (OA), which appears in EEG randomly as a big pulse. Based on discrete wavelet transform (DWT) and independent component analysis (ICA), a novel automatic method of OA removal, denoted as DWICA, was proposed. Firstly, DWT was applied to the recorded EEG and electrooculogram (EOG) to obtain multiple scale coefficients, and the combined coefficients were considered as the input for ICA. Secondly, the independent components were acquired based on FastICA algorithm with negentropy criterion. The angle cosine criterion was introduced to recognize ocular artifact component. Furthermore, the inverse algorithm of ICA was applied to project the independent components without OA to original electrodes. Finally, the EEG were reconstructed using the inverse algorithm of DWT, and then the pure EEG were obtained. Experimental results show that DWICA is preferable in automatic removal of OA. The method provides a new idea for on-line preprocessing of EEG signals.