Classification of Drivers’ Conditions

Drowsiness and inattention are a major cause for severe road accidents. The relationship Driver-Vehicle-Environment can be interpreted as a control loop in which a drowsy or inattentive driver controls worse than a driver in good condition. The quality of the control loop is reflected by specific patterns in the steering wheel angle, in the lane keeping, in the eye blinking behaviour, etc. These patterns can be detected by appropriate signal processing and classification methods. Such systems must be adaptive to the driving behaviour of every individual driver type and every road condition. To distinguish drowsiness related patterns from traffic- or road condition related patterns, it is furthermore necessary to detect crosswind, rough roads, road warping, vehicle operation, driving dynamics etc. The time of day and driving time since the last break can furthermore be used as a-priori knowledge. Sensor fusion methods can be applied to merge signals from different sources with different characteristics.

The research activities at the Chair of System Theory and Signal Processing in this area are the feature extraction and classification of drowsiness from the steering wheel angle, lane keeping and blinking behaviour and the fusion of different signals with different properties.