Modeling Body Motion Posture Recognition Using 2D-Skeleton Angle Feature

This paper proposed a method of human motion posture recognition using the angle characteristics of each branch. The coordinates of skeleton intersection and endpoint is extracted from the skeleton of human body image so as to calculate the branch angle parameters of head, hands and feet in human skeleton. In order to realize the matching identification of the series traffic command gestures, the method combined with SVM and template matching is used to classify and recognize the static human skeleton angle parameters. The method takes the effect on experiment of the resolution of the image and different size of human body into account, and the recognition rate can reach 93%, the recognition speed is 0.273s. This method also offers the theory and technology foundation in human body, pattern recognition and artificial intelligence field.