Table tennis coaching system based on a multimodal large language model with a table tennis knowledge base

Abstract

Table tennis is one of the most popular sports in the world, and it plays a positive role in the overall development of people’s physical and mental health. This study develops an AI table tennis coaching system using a Multimodal Large Language Model with a table tennis knowledge base, aiming to provide precise training guidance and match strategies for table tennis beginners. Method: By using visual recognition technology, motion capture technology, and advanced multimodal large language models with a comprehensive professional table tennis knowledge base, the system accurately identifies common errors made by beginners and provides targeted training guidance. Result: The AI Table Tennis Coaching System demonstrates high accuracy in identifying mistakes made by beginner players, particularly in recognizing arm-related errors and racket-related errors, with accuracies reaching 73% and 82% respectively. Conclusion: The system operates at low costs, is easy to deploy, and offers a high cost-performance ratio, providing effective technological support for table tennis teaching and training. The AI table tennis coaching system is expected to play a significant role in enhancing training efficiency, promoting athlete skill improvement, and popularizing the sport. Future research will focus on improving the accuracy of footwork recognition in AI table tennis coaching systems and expanding their capability to provide training guidance for high-level athletes, thereby promoting the overall advancement of table tennis.