A Framework for Human Identification using Video Sources

This paper presents a view invariant approach to recognize humans when engaged in some activity. Utilizing a particle swarm optimization (PSO) and linear discriminant analysis (LDA) based algorithm, an unknown movement is first classified, and, then, the person performing the movement is recognized from a movement specific person classifier. An un-calibrated multi-camera setup is used to capture the human body from different viewing angles. In the case of continuous person identification in videos depicting a person performing multiple instances of multiple activity classes, a sliding window of overlapping video segments consisting of video frames is used and person identification is performed at each window position Person identification, activity recognition, and viewing angle specification results are obtained for all the available cameras independently. Human identification performance of the proposed scheme is found to be quite good when tested on publically available databases.