Human pose discrimination using human bone structure features

Our goal in this paper is to develop a novel 2d human pose discrimination algorithm. With the speedy development in visual monitoring and internet, many images are needed to be coped with so that we can recognize the human poses that should be filtered or tracked. In this paper, we propose a novel human pose discrimination algorithm, in which three steps are utilized to access this goal, (1) finding seeds, (2)expanding pixel along the direction of one's limb bones, (3)constructing stickman, (4)pose matching with standard poses in library, (5)outputting discrimination results. Simulation is conducted to verify this algorithm, and we suggest the next step for the research. This algorithm is robust to scale and rotation.