Abstract: In order to reduce the false alarm rate of the advanced driver assistance system, a method for identifying driving intention is proposed by using the difference of “drivervehicleroad” parameters under different tasks. Experiments are carried out in driving simulator system, 1150 driving samples of 12 testees are recorded, and a driving intention recognition indicator system with 6 parameters are determined by comparing the sample difference of different driving intentions: lane keeping, lane change and overtaking. Using HMM and SVM cascade algorithm to establish driving intention recognition model. The results show that the correct recognition rate of driving intentions based on the algorithm reaches 9584%, obviously higher than that using HMM or SVM model alone, with an average single recognition time of 0017s, meeting the requirements of reaction time of driver to emergency events.

Key words: intelligent transportation, intention recognition, hidden Markov model, support vector machine；T-test