Multi-modal perception and decision-making system for smart sports classroom with cloud-edge-end collaboration

With the development of smart physical education classes, multimodal perception and personalized training have become the key to improving the quality of classes. This study proposes a smart physical education class system based on cloud-edge-end collaboration, which realizes data collection by end-side cameras, inertial sensors, heart rate monitors, etc., lightweight motion recognition and real-time feedback at the edge, as well as long-term personalized analysis and training prescription generation in the cloud. The system is experimentally verified in basketball, gymnastics and long-distance running classes. The average response time at the edge is 176–180 ms, the motion recognition accuracy is 89.4–93.2%, the stability index is 0.88–0.91, and the user satisfaction is 4.3–4.6 points. After enabling cloud-based personalized analysis, the motion recognition accuracy increased by approximately 3.4%, the response time decreased by 10.8 ms, and the user satisfaction increased by 0.6 points. These results indicate that the system can effectively optimize training effects and enhance personalized experience. The research results verify the advantages of the system in real-time performance, accuracy and personalized training, and provide a reference technical solution for the construction of smart physical education classes.