Video surveillance for elderly monitoring and safety

In this paper we propose a novel method to detect and record various posture-based and movement-based events of interest in a typical elderly monitoring application in a home surveillance scenario. Posture-based events include standing, sitting, bending/squatting, side lying and lying toward the camera. While movement-based events include running, jumping, active and inactive events. For posture classification, we use the projection histograms of foreground as the main feature vector. k-Nearest Neighbor (k-NN) algorithm and evidence accumulation technique is proposed to infer human postures. With this technique, we have achieved a robust posture recognition rate of above 90% and a stable classifier's output. Furthermore, we use the speed of fall to differentiate real fall incident and an event where the person is simply lying without falling. On the other hand, time series signal change detection techniques are used for movement classification task. The accuracy obtained for movement-based events detection is above 90%.