Abstract
The purpose of our research is to develop a sensor fusion technology that
monitors driver´s behavior to detect a transition of a driver's state into one that may possibly
lead to an accident-prone driving condition, and provide the driver with appropriate assistance
or support to enhance safety in a proactive and situation-adaptive manner.
Driver´s inattentiveness may result from driver's over-trust in or reliance on a driving support
system, occurrence of external distracting events, and lack of vigilance against traffic
conditions. This paper distinguishes three types of phenomena that may be caused by the
driver´s inattentiveness: (a) performing actions unnecessary for driving, or taking
inappropriate driving postures, (b) distraction due to some cognitive tasks, and (c) sleepiness
due to monotonous or extremely peaceful traffic conditions. Note here that type (b)
phenomena may be observed when a driver is required to perform a cognitive task, such as
communication with the company on urgent business matters, which usually imposes high
mental workload on the driver. We investigated how the measurement data may be affected by
the type of the inattentiveness.
The main results are: For detection of type (a) phenomena, it has been found that tactile
sensor information on the load center position (LCP) on the driver´s back is useful. By
analyzing dynamical changes of the LCP on the back, some sorts of driver's motions
irrelevant to the driving task could be detected. The type (b) phenomena can be detected
through monitoring and analyzing the movement of LCP on the seat or the eye movement.
It is true, however, that a detection algorithm based on single sensor information sometimes
failed to yield a correct detection. We are currently developing sensor fusion algorithms to
improve the rate of correct detection as well as to reduce the rate of false alarms.
Keywords - Driver monitoring, Complacency and over-trust, Distraction, Hypo-vigilance,
Sensor fusion