Driver hypovigilance diagnosis using wavelets and statistical analysis

This paper presents a driver vigilance diagnosis system based on the analysis of measured signals with wavelets and the computation of their statistical characteristics inside variable length time-windows. Such a system should characterize the driving mode of an unspecified vigilant driver in order to detect a modification of the way of driving due to the fall of vigilance. The signals that we assume to be able to characterize the driving mode are: the position of the vehicle inside the traffic lane, also called lateral position, the steering wheel angle and vehicle speed. This study has been performed with data collected in real traffic conditions. To determine the driving characteristics of vigilant driver we have selected few driving sequences where both the EEG analysis and the driver self-evaluation indicate the driver was vigilant. Then, we qualify the relevance of our system, particularly the choice of the variable used for the diagnostic, by comparing the diagnostic produces by our system with the physiological state of the driver based on EEG analysis and the driver self-evaluation.