Driver Drowsiness Detection using High Resolution Spectral Analysis of EEG

Details

18:15 - 20:15 | Mon 5 Mar | Caribbean ABC | MoPO.59

Session: Poster Session # 1 and BSN Innovative Health Technology Demonstrations

Abstract

In this study we aim to investigate driver's drowsiness impact on high resolution spectral power of the Electroencephalogram (EEG) signal in partially sleep-deprived driver while performing a simulated driving task. We implement a wavelet packet-based decomposition technique to decompose the EEG signal to high resolution spectral sub-bands. We investigate the drowsiness impact on its spectral power by comparing spectral power of all decomposed levels and sub-bands of EEG signal before and after drowsiness. Also we extract other features from all sub-bands and compare them to investigate the most informative and affective sub-bands of EEG signal from drowsiness. For this purpose, 5 subjects participated in driving simulator experiment while recording their driver behavior via EEG, motion sensors, ECG, and vehicular software data. The subject drivers went through 40-50 minutes of driving experiment under predefined scenario. Our observation illustrates the correlation between spectral power with the Karolinska Sleepiness Scale. Our experimental results show spectral power increment for lower sub-bands (less than 16 Hz) when driver is getting drowsy.