Performance Enhancement of Complete Ensemble Empirical Mode Decomposition (CEEMD) - Independent Component Analysis (ICA) In Ocular Artifact Removal

Abstract

Inaccuracies in removing Ocular Artifact (OA) in EEG signals can cause loss of information in EEG signals. Several previous methods utilize the feature of OA to detect OA and then remove all the components contaminated by OA. Those processes cause high errors in the OA removal results. A high error indicates that any information in EEG signals is lost during the OA removal process. The purpose of this research is to develop combination methods of CEEMD and ICA in order to improve performance in eliminating OA without omitting information in EEG signals. In its implementation, the combination of CEEMD and ICA are enhanced by the entropy method to find out the existence of OA, and Teager Kaiser Energy Operator (TKEO) method to measure the OA energy. Both ways can make it easier to detect and eliminate the OA by using the modified z-score method. The result of the test by using Relative Error (RE) indicates that the proposed method successfully eliminates OA with the average error of 0.4365 for Dataset A, 0.4597 for Dataset B, and 0.5337 for Dataset C. Thus, the results show the proposed method is able to remove OA without interfering information of EEG signal.