Detection of Driver’s Drowsiness Using New Features Extracted From Heart Rate Variability Signal

Abstract

As we may find in news related to road fatalities, we see more or less one third of these fatalities are because of drowsy driving or fatigue of drivers. Many researchers had investigations in detection of drowsiness of the drivers using biological signals e.g. ECG, EEG, EOG and image processing of drivers' faces or information shows how driving is going on. Using HRV signal extracted from one lead ECG or pulse oximetry signal is very simple and easy, but almost in all of these researches common features like time and frequency features and nonlinear features have been used. In this study, we used common features extracted from new signals in addition to features extracted directly from HRV. We first extracted HRV signal from ECG and then we construct Poincare map. From Poincare map we extracted two new signals. From these new signals we extracted some frequency and nonlinear features and then we used normal classifiers and reached to a higher sensitivity than before. Sensitivity up to 85% is usual in recent papers. We reached to a sensitivity 91.5 when we used selected features from both HRV and New signals.