Scholarlink

This paper presents an advanced system for automated pose creation, recognition, and correction, specifically designed to accommodate individual health conditions through the integration of the MediaPipe framework. While pose recognition technologies have significantly advanced human–computer interaction, virtual reality, and healthcare by enabling precise analysis of body movements, current solutions often lack the adaptability required for personalized feedback and the dynamic handling of new poses. Leveraging MediaPipe to extract 33 keypoints from real-time webcam input, the proposed system delivers immediate and accurate posture feedback during yoga and fitness exercises, achieving an overall accuracy of 85%. Unlike traditional approaches that rely on static datasets, this solution dynamically introduces new poses, thereby enhancing flexibility in fitness routines. Powered by machine learning techniques and an artificial neural network (ANN) architecture, the system refines feedback mechanisms based on user progress. Furthermore, it offers potential integration with wearable technologies for continuous health monitoring. This approach supports diverse applications in health improvement, physical therapy, fitness training, and immersive digital experiences.

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

Real-Time Motion Tracking and Exercise Pose Correction Using Media Pipe

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rishav Sinha,
  • Preet Kanwal

摘要

This paper presents an advanced system for automated pose creation, recognition, and correction, specifically designed to accommodate individual health conditions through the integration of the MediaPipe framework. While pose recognition technologies have significantly advanced human–computer interaction, virtual reality, and healthcare by enabling precise analysis of body movements, current solutions often lack the adaptability required for personalized feedback and the dynamic handling of new poses. Leveraging MediaPipe to extract 33 keypoints from real-time webcam input, the proposed system delivers immediate and accurate posture feedback during yoga and fitness exercises, achieving an overall accuracy of 85%. Unlike traditional approaches that rely on static datasets, this solution dynamically introduces new poses, thereby enhancing flexibility in fitness routines. Powered by machine learning techniques and an artificial neural network (ANN) architecture, the system refines feedback mechanisms based on user progress. Furthermore, it offers potential integration with wearable technologies for continuous health monitoring. This approach supports diverse applications in health improvement, physical therapy, fitness training, and immersive digital experiences.