Eye Blink Detection Using a Support Vector Machine Classifier

Abstract

In the last years, several applications for Brain Computer Interfaces (BCI) have been proposed, based on different kind of EEG features (e.g., mental states, cognitive load, alertness and eye blinks). Since eye blinks are considered artifacts for EEG when other EEG features are analyzed, many studies are focused on the detection of spontaneous eye blinks for removing. Only a few papers have investigated voluntary eye blinks classification, and most of them using different sensing techniques to EEG (Electrooculogram (EOG), Magnetic Resonance Imaging (MRI)). Toward development of a BCI application, the aim of this paper is to classify intentional eye blink events from EEG signals, to employ them as command controls in BCI applications, for people with special needs. In this paper, we trained a Support Vector Machine Classifier based on statistical features. The dataset is acquired using the low-cost Emotiv EPOC headset, using only a single EEG electrode through an experiment with a visual marker for 12 subjects.