Can variations in visual behavior measures be good predictors of driver sleepiness? A real driving test study

ABSTRACT

Objective: The primary purpose of this study was to examine the association between variations in visual behavior measures and subjective sleepiness levels across age groups over time to determine a quantitative method of measuring drivers' sleepiness levels.

Method: A total of 128 volunteer drivers in 4 age groups were asked to finish 2-, 3-, and 4-h continuous driving tasks on expressways, during which the driver's fixation, saccade, and blink measures were recorded by an eye-tracking system and the subjective sleepiness level was measured through the Stanford Sleepiness Scale. Two-way repeated measures analysis of variance was then used to examine the change in visual behavior measures across age groups over time and compare the interactive effects of these 2 factors on the dependent visual measures.

Results: Drivers' visual behavior measures and subjective sleepiness levels vary significantly over time but not across age groups. A statistically significant interaction between age group and driving duration was found in drivers' pupil diameter, deviation of search angle, saccade amplitude, blink frequency, blink duration, and closure duration. Additionally, change in a driver's subjective sleepiness level is positively or negatively associated with variation in visual behavior measures, and such relationships can be expressed in regression models for different period of driving duration.

Conclusions: Driving duration affects drivers' sleepiness significantly, so the amount of continuous driving time should be strictly controlled. Moreover, driving sleepiness can be quantified through the change rate of drivers' visual behavior measures to alert drivers of sleepiness risk and to encourage rest periods. These results provide insight into potential strategies for reducing and preventing traffic accidents and injuries.

KEYWORDS:

• Visual behavior measures
• subjective sleepiness level
• Stanford Sleepiness Scale
• Pearson correlation
• 2-way repeated measures ANOVA
• regression model

Acknowledgments

The authors acknowledge the Shandong Research Institute of Communications, Shandong Jiaotong University, and volunteers for providing cooperation in real driving test. We acknowledge the editors and 2 anonymous reviewers for their detailed comments and suggestions for improving the article, without which this work would not have been possible.

Funding

This research is financially supported by the National Natural Science Foundation of China (No. 51208051).

Compliance with ethical guidelines

This research was conducted in full compliance of the laws and after obtaining approval from each individual driver who took part in the real driving test to collect his or her eye movement and sleepiness awareness data.