Blind Source Separation Methods Applied to Muscle Artefacts Removing from Epileptic Eeg Recording: A Comparative Study.

Archive ouverte

Kachenoura, Amar | Safieddine, Doha | Albera, Laurent | Birot, Gwénaël | Wendling, Fabrice | Senhadji, Lotfi | Merlet, Isabelle

Edité par CCSD -

International audience. Electroencephalogram (EEG) recordings are often contaminated with muscle artifacts. These artifacts obscure the EEG and complicate its interpretation or even make the interpretation unfeasible. In this paper, realistic spike EEG signals are simulated from the activation of a 5 cm2 epileptic patch in the left superior temporal gyrus. Background activities and real muscle artifacts are then added to the simulated data. We compare the efficiency of Empirical Mode Decomposition (EMD), Independent Component Analysis (ICA) and Blind Source Separation based on Canonical Correlation Analysis (BSS-CCA) to remove muscle artifacts from the EEG signals. The quantitative comparison indicates that the EMD approach exhibits a better performance than ICA and BSS-CCA, especially in the case of very low Signal to Noise Ratio (SNR).

• Description
• Sujet/Public
• Outil
• Outil d'anticipation
• Mémoire et thèse
• Gestion