Driver fatigue discrimination method based on visual features

Abstract

Driver fatigue is a major cause of road accidents and kills tens of thousands of people per year. Eye movement and visual features are considered to be affected by driver fatigue. In this paper, a 120min driving simulation test was designed, and the Stanford Sleepiness Scale (SSS) subjective fatigue values and eye characteristics of 16 drivers were collected. Subjective fatigue values were divided into three grades: awake, awake-fatigue and fatigue. From the perspective of visual characteristics, the differences in saccade, blink, and fixation between awake and fatigue states were analyzed. Finally, saccade, blink, fixation, and combination methods to discriminate driver fatigue were compared according to the classification accuracy. The results show that whether drivers are awake or fatigued, their fixation duration is longer than blink duration, and blink duration is longer than saccade duration. Under driver fatigue, saccade duration and fixation duration decreased significantly, while blink duration increased significantly. The classification accuracy of the proposed method in this paper is more than 80%, indicating that visual features have a certain application value in the detection of driver fatigue. The results of this study can provide some suggestions for driver fatigue prevention and control and also help to ensure traffic safety.