Intelligent Sports performance Scoring and Analysis system Based on Deep Learning Network

Abstract

In the past, sports analysis is always conducted by specialized devices like sensors or depth camera, which are hard to promote. In this essay, we use computer vision technology to establish a sport detection and analysis system with the use of an RGB camera and deep learning models. It can recognize the posture from an RGB image, make an overall evaluation, and return the details real-time.The RGB image will be processed by two sub-systems sequentially: Human detection system and Performance analysis system. The human detection system is used to detect the user's posture, while the performance analysis system is designed to return feedback on the details of the user's posture.The results of the system are satisfactory:1)In the human detection system, the overall accuracy can reach more than 99%, meaning that almost the 4 actions can be classified accurately.2)As for the performance analysis system, the score and details can be returned correspondingly based on tiny differences within the centimeter level.3)The delay of the whole system is within milliseconds, meaning that the system can achieve real-time detection and analysis.In this essay, golf driving is taken as an example to illustrate how the system works. Also, the system can be promoted to other sports.