Falls are a major problem for elderly living independently. The World Health Organization reports that injuries due to falls are the third most common cause of chronic disability. We propose an automated monitoring system that identifies faces in a given area, collects data such as speed of movement, and triggers an alarm if these data suggest the person has fallen. Our system does not suffer the problems with the existing commercial devices such as social alarms, e.g., a wristwatch with a button that must be activated by the subject and wearable fall detectors comprising accelerometers and tilt sensors.