Abstract: In order to effectively recognize the driving intention triggering on the vehicle safety assistant system, we studied the relationship between the driving intention and driving behavior. Based on hidden Markov model (HMM) hierarchical and sequential characteristics, we established a double layer HMM model, the behavior layer and the intention layer, respectively, used for recognizing driving behavior and driving intention. After using the experimental data processed, we carried out off-line training for each identification model of the double layer HMM model. Moreover, through comparing the offline validation results of different model parameters, we selected the appropriate parameters for each model, which can achieve the best recognition results. The research results show that the intention layer can identify long and complex driving intentions in real time. It is concluded that this research has successfully built the online recognition model of driver’s intention.
