Posture and Activity Recognition Using Projection Histogram and PCA Methods

In this paper, we propose an intelligent surveillance system which can detect human, recognize certain postures and activities, and send out informing SMS messages when abnormal activities are detected. Nine static postures can be recognized using universal eigenspace technique and projection histogram matching method. After recognizing static postures, Hidden Markov Models are used to recognize human behaviors. If the behavior is regarded as an abnormal activity, an informing message will be sent out.