A fully automatic ocular artifact removal from EEG based on fourth-order tensor method

Purpose The aim of this paper is to propose a fully automatic system using the underdetermined blind source separation (UBSS) method and kurtosis to remove ocular artifacts (OAs) from scalp electroencephalogram (EEG). Methods The fully automatic system about OAs rejection is devised with the fourth-order tensor method (FOOBI). Firstly, the FOOBI method decomposes multiple EEG channels into a relative large number of source components. The kurtosis value is used to identify the ocular components in these source components. Then, the free-ocular sources components are reconstructed to EEG without OAs. Results The simulations show that the FOOBI method can completely separate the ocular signals from the observed signals. The data that have got rid of the OAs are used to classify with the epileptic EEG. The classification accuracy acquired by FOOBI method is better than the independent component analysis (ICA). Conclusions The results inferred that the FOOBI method can not only completely remove the OAs from the observed signals but also preserve an amount of useful information of EEG. Compared with the ICA method, this fully automatic system is more suitable to remove OAs.