Multivariate Fast Iterative Filtering Algorithm for Denoising of Ocular Artifacts in EEG

The elimination of artifacts including ocular movements in multichannel long-standing scalp EEG recordings has been the goal of extensive research with the aim to distinguish epileptic seizure episodes in EEG signals. The current research introduces a method for effectively eliminating ocular artifacts (OA) from extended EEG recordings by employing multivariate fast iterative filtering (MVFIF) in conjunction with a total variation (TV) denoising technique. The current approach initially employs the MVFIF for signal decomposition and then filters the selected noisy IMFs with a TV denoising scheme. The evaluation of the denoising impact involves mean absolute error (MAE), the power spectral density distortion (PSDD), and mutual information (MI) metrics. Applied to actual EEG data, the proposed methodology successfully eliminates OA components and yields a minimal MAE of 0.0020.