Drowsy Driver Detection Based on Steering Wheel and Lateral Position Variables Using Driving Simulator

Drowsy driving is a main cause of severe accidents. Drowsiness is responsible for 30% to 37% of fatal road accidents in Iran. In this paper, driver drowsiness was detected based on features related to steering wheel angle and lateral position of the vehicle. Vehicle and the virtual road data were used to extract drowsiness features. Experimental results using a driving simulator were presented. Participants were 21 to 28 year-old males with a high tendency to sleep (Epworth Sleepiness Scale ≥ 10). The subjects had to drive a lane keeping scenario on a long and monotonous virtual road in both drowsy and alert states. The drowsiness criteria were validated using Karolinska Sleepiness Scale (KSS) and video rating based on KSS measurements. The results illustrated that steering wheel angle phase diagram (Ellipse criterion), standard deviation of steering wheel angle, and mean and standard deviation of lateral position of the vehicle are highly correlated with drowsiness. The accuracy of the diagnosis was 77% for Ellipse criterion, 76% for standard deviation of steering wheel angle, 67% for standard deviation of lateral position, and 65% for mean of the lateral position with few false positive alarms.