A rapid recognition of athlete's human posture based on SVM decision tree

Abstract

In order to solve the problems of low recall rate of human posture data collection results, low recognition rate and long recognition time in traditional recognition methods, a rapid recognition method of athlete's human posture based on SVM decision tree was proposed. The Kinect sensor is used to collect the athlete's human posture data, and the mixed Gaussian background modelling method is used to segment the collected athlete's human posture image. Scale normalisation is performed on the segmented images, and a star model is used to extract the pose features of athletes' bodies. According to the characteristics of human posture, the SVM decision tree is used to classify and identify the human posture of athletes. The experimental results show that the maximum recall rate of this method is 98%, the minimum value is 93%, the recognition rate is above 97.2%, and the average recognition time is 0.62.