Pedestrians account for a significant ratio of traffic fatalities; as a result, research on methods of reducing vehicle-pedestrian crashes is of importance. In this paper, we describe a system architecture that allows the use of vehicle-to-pedestrian (V2P) communication as a means of generating situational awareness and eventually predicting hazards and warning drivers and pedestrians. In contrast, vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication for safety applications, V2P has not received much attention. One major reason for this lack of attention had been the unavailability of communication mechanisms between pedestrians and vehicles. Recent advances in enabling Wi-Fi and dedicated short range communication (DSRC) based communication using smart-phones is changing this picture. As a result, V2P communication can be considered as a possible solution. In this paper, we first investigate the feasibility of V2P crash avoidance applications, considering the fact that pedestrian movement dynamics, response times, and crash scenarios are different from V2V scenarios. We then propose an architecture for V2P crash avoidance systems. The architecture has vehicle side and pedestrian side applications. Each of them communicates its current state information such as position and time to the other device. Using this information, crash threats can be predicted and an advisory or imminent collision warning will be generated. Finally, we demonstrate and evaluate a developed V2P crash avoidance system based on this architecture.