Intelligent Human Body Modeling and Human Pose Classification Using Neural Networks

With increased demand for security systems, video analysis plays an important role in surveillance systems. This paper describes an intelligent visual surveillance system that detects the presence of the human body parts and actions. The motion segmentation is achieved by background subtraction algorithm. The body parts such as head, torso and both hands are detected individually from the segmented image. First, the head region is identified using the Vertical Projection Histogram (VPH) algorithm. K-means clustering can be used to segment the skin regions of the monocular video sequences. For the detected skin regions, 2D modeling has been implemented for pre-defined poses. The human pose classification can be done by Feed forward Back propagation neural networks algorithm. In pose classification, ten poses are considered. The proposed work is used to detect the skin regions even if the person wears half sleeved or full sleeved dress. Finally, the efficiency of pose classification is also analyzed