Occupant restraint systems are developed based on some baseline experiments, while these experiments can only represent a small part of various accidents, so the current design of occupant restraint systems cannot provide the optimum protection effects for occupants during the majority of accidents. This study presents an approach to predict occupant injury before the collision happens, so that the occupant restraint system can be adjusted to the optimum parameters aiming at the imminent vehicle-to-vehicle frontal crash. The approach in this study takes advantages of the information from pre-crash systems, such as the time to collision, the relative velocity, the frontal overlap of the ego vehicle during the collision, the size of the vehicle in the front and so on. The information acquired and the basic crash test results can be integrated to predict a simplified crash pulse, and the injury of the occupant in the ego vehicle can thus be predicted using this crash pulse. The approach is verified by some vehicle-to-vehicle crash tests, and the results show that the prediction of occupant injury by this approach is close to the dummy injury responses in the tests.