Driver Behavior Response to Drowsiness Alarming at Different Levels

To identify the timing of drowsiness driving warning is the key issue and a bottleneck of onboard drowsiness driving warning technology. Finding a rationale for warning timing using driver’s driving behavior response feature is an innovation. Therefore, after conducting a driving simulator experiment under the influence of drowsiness alarming, eye movement index, the percentage of eyelid closure (PERCLOS), and vehicle lateral position indexes (standard deviation of lateral position, average of lateral position, area of line exceeding) were recorded. Then, the differences in driving behavior between the pre-warning and post-warning in 15 seconds for each warning were compared using the paired Wilcoxon signed-rank test. The results demonstrated that under the drowsiness classification criterion, after the normal level of drowsiness warning, the mean of the standard deviation of lateral position and the mean of the area of line exceeding significantly dropped down by 0.129 1 and 8.574 4 respectively; after the serious level of drowsiness warning, although the mean of PERCLOS decreased by 0.044 9, the standard deviation of the lateral position and area of line exceeding did not significantly change and that the driver should stop and rest immediately.