Distraction and drowsiness – a field study. Technical report

VTI
April 2009

The main goal of the distraction and drowsiness field study was to evaluate a system for detecting driver distraction and drowsiness. This report focuses on the system implementation and the algorithms detecting distraction and drowsiness. A previous report dealt with the background of driver distraction, the results of the study are presented in a further report.

A vehicle was instrumented with video cameras, an automatic eye tracker and GPS receivers. Further data were read from the CAN bus of the car. The data were logged continuously with high frequency as long as the ignition was on. The log system operated autonomously and was switched on with the turning of the ignition key. Seven participants drove the vehicle during one month each. During the first ten days the distraction and drowsiness warning system was deactivated, in order to collect a behavioural baseline. After this the warnings were activated, such that the driver received distraction warnings in form of a vibration in the seat when the algorithm determined that they had looked away from the forward roadway too much. A separate algorithm judged whether the drivers were drowsy or not. Three drowsiness levels existed, which led to three different warnings with increasing intensity for increasing drowsiness. The eye tracking system installed in the car afforded real time eye tracking, which is necessary for giving glance direction based or blink duration based warnings in real time.

The participants filled in questionnaires about their driving habits, their attitudes about driver distraction and drowsiness, and their expectations towards the warning system as well as their experiences with the system. A number of questionnaires were administered on three occasions during the course of the study.

No major problems were encountered during the field study, but a number of smaller problems had to be solved. However, in the end of the data collection period the computer installed in the car became more and more unstable, which led to increased data loss. Again, piloting was shown to be essential, as well as a clear delimitation of the goals and hypotheses of the project.

The obvious potential of the distraction detection system was pointed out by the positive comments of the participants, but having a system which is reliable in all situations and for all drivers is still a difficult task.

Acceder al documento