The work aims to explore the impact of the eHMI of the autonomous vehicle, the road traffic environment and individual characteristics on the pedestrian's crossing decision time when pedestrians interact with the autonomous vehicle, and determine the optimal form of the eHMI. The experimental scenario of pedestrian-vehicle interaction was constructed by Unity 3D, the volunteers were recruited to participate in the simulation experiment, the crossing decision time and behavior data of pedestrians under different conditions were collected and the statistical analysis was carried out by the survival analysis model. The multi-modal (visual + auditory + physical) eHMI significantly reduced the pedestrian's crossing decision time, and its crossing efficiency was 3.907 times higher than that of the single physical modal interface. The pedestrian's age was proportional to the decision time. Compared with the young people, the crossing efficiency of the middle-aged was 0.532 times that of the young people, while the crossing efficiency of the elderly was further reduced to 0.316 times that of the young people. In addition, the vehicle approaching from the near lane also significantly shortened the pedestrian's decision time. When the vehicle approached from the far lane, the pedestrian's crossing efficiency was only 70.1% of that when the vehicle approached from the near lane. The gender has no significant effect on the pedestrian's crossing decision time. The eHMI of the autonomous vehicles, the pedestrian's age and the vehicle approaching lane all have significant impacts on pedestrians' crossing decision time. The highest crossing efficiency can be achieved when pedestrians face the eHMI of visual + auditory + physical information.