A Gaze-Based Driver Distraction Warning System and Its Effect on Visual Behavior

Driver distraction is a contributing factor to many crashes; therefore, a real-time distraction warning system should have the potential to mitigate or circumvent many of these crashes. The objective of this paper is to investigate the usefulness of a real-time distraction detection algorithm called AttenD. The evaluation is based on data from an extended field study comprising seven drivers who drove on an average of 4351 ± 2181 km in a naturalistic setting. Visual behavior was investigated both on a global scale and on a local scale in the surroundings of each warning. An increase in the percentage of glances at the rear-view mirror and a decrease in the amount of glances at the center console were found. The results also show that visual time sharing decreased in duration from 9.94 to 9.20 s due to the warnings, that the time from fully attentive to warning decreased from 3.20 to 3.03 s, and that the time from warning to full attentiveness decreased from 6.02 to 5.46 s. The limited number of participants does not allow any generalizable conclusions, but a trend toward improved visual behavior could be observed. This is a promising start for further improvements of the algorithm and the warning strategy.