Detecting Suspicious Motion with Nonimaging Sensors

Automated distributed sentry systems need only detect suspicious behavior. Microphones and infrared detectors may suffice, as well as being simpler and cheaper than cameras while reducing invasion of privacy. We report on a new approach to analyzing data from acoustic sensors and binary infrared and magnetic sensors towards detecting changes in the velocity vectors of walking people. We show that changes in speed and direction can be detected directly by simple mathematical calculations from even imprecise sensing data. We show methods for detecting both locally and globally suspicious behavior. We present results of experiments supporting our approach, and discuss design of a.