Drowsiness Detection with a Limited Number of EEG Physiological Signals

Abstract

A variety of studies have already been carried out to try to discriminate the different stages of alertness of a human subject. The purpose of this paper is to propose a method for detecting drowsiness in drivers by adopting a real-time analysis of EEG activity. First, we introduce our database collected at the Technology and Medical Imaging (TIM) laboratory of the University of Monastir, Tunisia. Second, we propose a method for the detection of the decrease of vigilance from a single EEG channel. This method, based on the SVM classifier, was tested on the collected database and allows to detect drowsiness results up to 91.39% in terms of accuracy.