While many of today's devices feature built-in video cameras, they do not have an embedded computer vision system capable of recognizing the movement of individual objects within larger gatherings or groups. This would be essential, however, to perform demanding tasks such as guiding robots, measuring traffic or tracking people during videoconferences. The key is spatiotemporal movement recognition. The project is to segment videos in time and space and will be developed into working prototypes, initially for desktop computers, then also for handheld-sized devices. The prototypes target a variety of applications. One example, people-counting in crowds, is intended to provide valuable information to architects and urban planners who study crowd density and resolve bottlenecks of pedestrian traffic at different times of the day. Such information is also important when planning large sporting events where safety criteria or public transit frequency must be considered. Embedded systems would be mounted near traffic signs, for example, and periodically send back information about how many people are going which way, without affecting the privacy of individuals or violating data protection provisions.