Application of multivariate empirical mode decomposition and canonical correlation analysis for EEG motion artifact removal

Electroencephalogram (EEG) is a tool to record the electrical activity of brain. It often records artifacts which are the electrical activities originating from sites other than brain. The presence of artifacts is undesirable as it increases the probability of misinterpretation that may result in adverse clinical consequences. In this paper, we present a novel method of motion artifact removal from the EEG signals. The proposed method employs multivariate empirical mode decomposition (MEMD) with canonical correlation analysis (CCA) for the removal of motion artifacts from EEG signal. The mode alignment property of the multivariate empirical mode decomposition is used in EEG signal analysis where a similarity between different channels is the key to decode the signals. When compared with the results of existing methods for motion artifacts removal on the same data set, the MEMD-CCA is shown to perform better with 16 % increase in the percent artifact removal. Unlike the other competing methods, this method is totally algorithmic and does not require any manual intervention.