Abstract: To accurately evaluate the level of driver's risk perception during driving, data on driving behavior, electrophysiological status and eye movement of driver's response to different traffic flow conditions, intersection geometry and sudden events were collected, together with the data on vehicle performance, vehicle behavior and the interaction with other vehicles. The risk assessment indicators were analyzed from the perspectives of driver's physiological and physical indicators. Based on the risk perception quantification, scenario extraction and behavior analysis, an evaluation index system was established. Then, by using the HMM, quantitative values of risk perception were predicted, driving behavior parameters under different risk perception quantitative values were analyzed, and an observation sequence of risk perception quantization values was finally obtained. The validity of the model was verified. The results show that the HMM model can predict the quantified value of driver's risk perception with an accuracy over 85%.

Key words: traffic engineering, risk perception, physiological behavior, vehicle behavior, hidden Markov model(HMM)