The Application of Multivariate Empirical Mode Decomposition with Canonical Correlation for EEG Artifact Removal

Includes bibliographical references (pages 38-30).A variety of algorithms have been developed for removing unwanted artifacts from physiological signals. However, artifact removal continues to be an open research problem as there is currently no single best method that is efficient and robust over a wide range of conditions. Empirical Mode Decomposition is a fully data driven method for the analysis of nonlinear and non-stationary real world signals. The Multivariate version of the EMD algorithm (MEMD) is used to find common oscillatory modes within multivariate data. This feature of the MEMD called mode-alignment, is used in EEG signal analysis where similarities between different channels is the key to decode the signals. In this thesis, we propose a novel method multivariate empirical mode decomposition with canonical correlation analysis (MEMD-CCA) for the removal of motion artifacts from electroencephalography (EEG) signal. The proposed technique is then compared with the existing competing methods for motion artifacts removal. The MEMD-CCA is shown to perform better. The computational cost of this method is higher but it is more efficient in removal of motion induced artifacts