Comparison of blind source separation methods for removal of eye blink artifacts from EEG

Abstract

Electroencephalography (EEG) recording are generally corrupted by eye blink artifacts. In this research work, blind source separation (BSS) based methods for removal of eye blink artifacts are presented. Two techniques, namely the Independent Component Analysis (ICA) and the Canonical Correlation Analysis (CCA) are investigated. The efficiency and performance of the BSS methods were compared between the two methods using simulated contaminated EEG data of three channels. ICA recovers the EEG signals with average correlation coefficients of 0.5185, 0.7906 and 0.8217 for EEG signals 1, 2, and 3, respectively. The ICA improves signal-to-artifact ratio (SAR) from -4.1395 to 5.9685. On the other hand, CCA recovers the EEG signals with average correlation coefficients of 0.5739, 0.8229 and 0.8427 for EEG signal 1, 2 and 3, respectively, and it improves SAR from -3.5709 to 7.6891. In addition, elapsed time is also investigated for both methods. Average elapsed time of 0.0114s and 0.0905s were computed for ICA and CCA, respectively. These simulated results demonstrate that CCA is more accurate and faster than ICA.