Statistical validation of physiological indicators for non-invasive and hybrid driver drowsiness detection system

Eugene Zilberg, Zheng Ming Xu, David Burton, Murad Karrar, Saroj Lal

Abstract


A hybrid system for detecting driver
drowsiness was examined by using piezofilm movement
sensors integrated into the car seat, seat belt and steering
wheel. Statistical associations between increase in the
driver drowsiness and the non-invasive and conventional
physiological indicators were investigated. Statistically
significant associations were established for the analysed
physiological indicators – car seat movement magnitude
and (electroencephalogram) EEG alpha band power
percentage. All of the associastions were physiologically
plausible with increase in probability of drowsiness
associated with increases in the EEG alpha band power
percentage and reduction in the seat movement
magnitude. Adding a non-invasive measure such as seat
movement magnitude to any combination of the EEG
derived physiological predictors always resulted in
improvement of associations. These findings can serve as
a foundation for designing the vehicle-based fatigue
countermeasure device as well as highlight potential
difficulties and limitations of detection algorithm for such
devices.

Keywords


driver fatigue, drowsiness, detection algorithm, seat movements, EEG, alpha activity

Full Text:

PDF PDF