Estimation of Driver's Arousal State Using Multi-dimensional Physiological Indices

The goal of our research is to develop a method to assess the arousal states using facial images of drivers. Multi-dimensional physiological indices are expected to be alternative external criteria of arousal states to manual coding of facial expression which require a lot of human resources. Changes in multi dimensional physiological indices (i.e., blink categories, skin conductance, EEG alpha wave, respiration, heart rate variability) depending on the arousal states defined by the combination of "arousal level" and the presence of "effort" to wake up were studied. Multiple linear regression analysis was also executed using one of the face scores ("arousal level" or "effort") as dependent variables and the physiological indices as explanatory variables. Relatively high multiple correlation coefficients were obtained, however, the number and combination of selected indices showed great differences between individuals. To obtain common equations is an issue in future.