Hybrid Method Combining VMD, Wavelets, and BSS for Efficient Removal of Ocular Artifacts in EEG Signals

The removal of ocular artifacts in EEG signals is an important challenge in all fields and especially in the field of brain-machine interface. This study proposes a hybrid methodology combining variational mode decomposition (VMD), discrete wavelet (DWT) and blind source separation (BSS) techniques. This combination is based on the decomposition of EEG signals into intrinsic modes using VMD, followed by an analysis with symlet wavelets. Finally, BSS algorithms were applied to separate independent sources and extract the signals of interest while preserving the relevant brain information. The performance of the algorithm was evaluated using correlation and Euclidean distance measures between the separated components and the reference signals (HEOG and VEOG). The results, obtained from 54 EEG recordings, show that this hybrid approach significantly reduces ocular artifacts. However, there are still limitations, especially when it comes to artifact separation accuracy in realtime application contexts. These results show the importance of combining multiple techniques in order to improve EEG signal processing.