Research and Development of Intelligent Recognition system for pull-up Action Norms Based on OpenPose

Pull-up is one of the common sports in national fitness and a compulsory item in the upper limb strength test for youth group required by the National Physical Exercise Standard Work Guidance Manual (2020 version). Currently, there is a lack of professional pull-up action specification recognition system, and there are problems such as unified counting standard and unadjustable difficulty level. The study uses computer vision technology based on deep learning to develop an intelligent recognition system for action specification, and proposes a difficulty-adjustable recognition algorithm model, which helps to improve the standardization level of pull-ups. The system adopts dual cameras to detect the key points of human joints by OpenPose, a human posture recognition framework, to judge the validity and normality of pull-up movements based on the key point positions and joint angle change characteristics, and to switch between different difficulty levels by adjusting the judgment conditions of movement norms.