Eye blinks artefacts detection in a single EEG channel

Abstract

In clinical conditions, the EEG measurements are mainly influenced by muscles and ocular movements, especially eye blinks. In this work, we present a new method to detect and reject eye blinks from a single-channel EEG signal. In an offline use mode, the proposed approach is based on statistical computations. It aims, in a first time, at estimating the pure EEG data interval. In a second time, we seek to improve the errors that may occur during the first step by using the Fisher-Snedecor test. By using ROC performance metrics, kappa coefficient and signal-to-artefact ratio (SAR), the proposed method is compared to an expert detection and to single-channel independent component analysis (SC-ICA), one of the widely used and robust methods for artefacts rejection. Experimental results show large values even when using the expert's annotation and the comparison to the SC-ICA method. This reflects the efficiency of the proposed method in detecting and rejecting blinks from a single-channel EEG signal.