We investigate the pupil response to hazard perception during driving simulation. Complementary to gaze movement and physiological stress indicators, pupil size changes can provide valuable information on traffic hazard perception with a relatively low temporal delay. We tackle the challenge of identifying those pupil dilation events associated with hazardous events from a noisy signal by a combination of wavelet transformation and machine learning. Therefore, we use features of the wavelet components as training data of a support vector machine. We further demonstrate how to utilize the method for the analysis of actual hazard perception and how it may differ from the behavioral driving response.
展开▼
