ABSTRACT：Drowsy driving is one of the main causes of traffic crashes, and it is a serious threat to road traffic safety. The effective early detection of a drowsiness state will help to provide a timely warning for drivers. The purpose of this study is to establish a model to detect the driver's drowsiness level by considering the individual differences (including driving behavior and eye characteristics) combined with the time cumulative effect of a driver's drowsiness. The driving behavior and eye movement data of 15 drivers were collected by combining driving simulator and eye-tracking system. The Karolinska Sleepiness Scale (KSS) was used to evaluate the driver’s subjective drowsiness state. Based on the individual differences of drivers and the time cumulative effect of drowsiness, a mixed-effect ordered logit (MOL) model considering the time cumulative effect (TCE) was established, since previous studies seldom consider the effect of drowsiness with time accumulation. The drowsiness degree of drivers was shown to increase with time. In order to study time effects on drowsy driving, a nondecreasing function of time was applied to the MOL model. The accuracy of the MOL-TCE model for the detection of drowsiness is 80.35%, higher than MOL model without considering time cumulative effect (67.75%). With increasing drowsiness, the standard deviation of lateral position and percentage of eyelid closure of drivers increased significantly. The results show that standard deviation of lateral position, percentage of eyelid closure and average pupil diameter have a significant effect on the detection of drowsiness and thus improves the accuracy of drowsy driving detection.
Xuxin Zhang, Xiaohan Yang, Xuesong Wang*, Chuan Xu. Driver Drowsiness Detection Using Mixed-Effect Ordered Logit Model Considering Time Cumulative Effect. Transportation Research Board 98th Annual Meeting, Washington D.C., USA, 2019. 1.13-17.