Scholarlink

This paper presents the development and implementation of YogaPose Tracker, an AI-driven posture analysis and training management application designed to enhance yoga practice through real-time feedback. The system provides accurate human pose estimation and posture correction by leveraging the MoveNet Thunder deep learning model and TensorFlow.js. The application runs on a Raspberry Pi-based setup equipped with a camera and display, enabling real-time analysis and feedback. It features a user-friendly interface with split-screen visual feedback, audio guidance, and session tracking to ensure an interactive and engaging training experience. Comprehensive testing demonstrated high accuracy in yoga pose detection, with a maximum recorded accuracy of 96.72%. The system’s adaptability was further validated in archery training, where it provided feedback on shooting posture, achieving archer accuracy levels of up to 96.15%. These results highlight the potential of AI-powered posture tracking beyond yoga, making it applicable to various sports and fitness activities. Future improvements include integrating multiple cameras, leveraging GPU acceleration for enhanced processing speed, and incorporating wearable sensors for a more comprehensive training experience. By combining AI, real-time feedback, and an intuitive interface, YogaPose Tracker advances digital fitness solutions and personalised coaching technologies.

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

AI-Driven Posture Analysis and Training Management Application

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Camelia Avram,
  • Alexandra Tăslăuan,
  • Dan Radu,
  • Adina Aştilean

摘要

This paper presents the development and implementation of YogaPose Tracker, an AI-driven posture analysis and training management application designed to enhance yoga practice through real-time feedback. The system provides accurate human pose estimation and posture correction by leveraging the MoveNet Thunder deep learning model and TensorFlow.js. The application runs on a Raspberry Pi-based setup equipped with a camera and display, enabling real-time analysis and feedback. It features a user-friendly interface with split-screen visual feedback, audio guidance, and session tracking to ensure an interactive and engaging training experience. Comprehensive testing demonstrated high accuracy in yoga pose detection, with a maximum recorded accuracy of 96.72%. The system’s adaptability was further validated in archery training, where it provided feedback on shooting posture, achieving archer accuracy levels of up to 96.15%. These results highlight the potential of AI-powered posture tracking beyond yoga, making it applicable to various sports and fitness activities. Future improvements include integrating multiple cameras, leveraging GPU acceleration for enhanced processing speed, and incorporating wearable sensors for a more comprehensive training experience. By combining AI, real-time feedback, and an intuitive interface, YogaPose Tracker advances digital fitness solutions and personalised coaching technologies.