Scholarlink

<p>The recent rise in the popularity of digital sports training tools has brought into focus the need to have intelligent, real-time, and personalized performance analysis. The traditional methods of coaching are not very scalable, flexible, or correct, particularly in remote or resource-constrained locations. The combination of Artificial Intelligence (AI) and Internet of Things (IoT) can be used to overcome these weaknesses by providing continuous surveillance and feedback. However, the challenges of low-latency data processing, scalable deployment, and personalization to an athlete are yet to be addressed. The paper suggests and analyzes a proposal of an AI-IoT integrated system, CoachXNet, which is proposed to deliver personalized digital training and instant corrective feedback in sports. It leverages deep learning algorithms based on hybrid algorithms to predict the pose and produce suggestions, and takes advantage of edge-cloud cooperation to balance the latency, scalability, and resource usage. In order to test the system, the SportsPose and AthletePose3D datasets were incorporated into the experiments. Motion capture was carried out through IoT-based wearable and vision-based sensors, and preprocessing and augmentation were applied to enhance model generalization. The low-latency inference was done via a server on the edge, but the model updates were done on the cloud resources on a large scale. The metrics used to evaluate were the accuracy of the action recognition, Mean Per Joint Position Error (MPJPE), the latency, and the efficiency of providing the feedback. CoachXNet performed better than the baseline frameworks with an accuracy of 94.1%, an MPJPE of 35.2&#xa0;mm, and an average end-to-end latency of 32 ms. Individualized training recommendations resulted in better outcomes of athlete performance by 18–23% than non-individualized training recommendations. CoachXNet shows that it is possible to implement an AI-IoT integrated feedback loop in sports training and increase accuracy, responsiveness, and personalization of the system significantly. The results indicate its capabilities in developing scalable athlete-centered digital coaching systems for the future generation of sports ecosystems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CoachXNet: An Artificial Intelligence and Internet of Things Integrated Platform for Personalized Training and Feedback in Digital Sports

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月19日

查看全文
  • Yongkang Wang,
  • Xuehan Feng,
  • Yujia Wu,
  • Xi Zhang,
  • Jian Wu

摘要

The recent rise in the popularity of digital sports training tools has brought into focus the need to have intelligent, real-time, and personalized performance analysis. The traditional methods of coaching are not very scalable, flexible, or correct, particularly in remote or resource-constrained locations. The combination of Artificial Intelligence (AI) and Internet of Things (IoT) can be used to overcome these weaknesses by providing continuous surveillance and feedback. However, the challenges of low-latency data processing, scalable deployment, and personalization to an athlete are yet to be addressed. The paper suggests and analyzes a proposal of an AI-IoT integrated system, CoachXNet, which is proposed to deliver personalized digital training and instant corrective feedback in sports. It leverages deep learning algorithms based on hybrid algorithms to predict the pose and produce suggestions, and takes advantage of edge-cloud cooperation to balance the latency, scalability, and resource usage. In order to test the system, the SportsPose and AthletePose3D datasets were incorporated into the experiments. Motion capture was carried out through IoT-based wearable and vision-based sensors, and preprocessing and augmentation were applied to enhance model generalization. The low-latency inference was done via a server on the edge, but the model updates were done on the cloud resources on a large scale. The metrics used to evaluate were the accuracy of the action recognition, Mean Per Joint Position Error (MPJPE), the latency, and the efficiency of providing the feedback. CoachXNet performed better than the baseline frameworks with an accuracy of 94.1%, an MPJPE of 35.2 mm, and an average end-to-end latency of 32 ms. Individualized training recommendations resulted in better outcomes of athlete performance by 18–23% than non-individualized training recommendations. CoachXNet shows that it is possible to implement an AI-IoT integrated feedback loop in sports training and increase accuracy, responsiveness, and personalization of the system significantly. The results indicate its capabilities in developing scalable athlete-centered digital coaching systems for the future generation of sports ecosystems.