Related Experiment Video
Updated: May 15, 2026
06:37
Artificial Intelligence-Based System for Detecting Attention Levels in Students
Published on: December 15, 2023
Application of deep learning based motion posture recognition in sports training
Yingying Shi1, Junyi Yu2, Yizhen Li2
1College of Physical Education, Chengdu Normal University, Chengdu, 611130, Sichuan, China. shiyingying198@outlook.com.
Scientific Reports
|May 13, 2026
Summary
A new deep learning model, Deep Dynamic Graph Attention Posture Recognition (DDGAPR), enhances athletic movement analysis. This AI-assisted approach improves sports training accuracy and personalization for better performance optimization.
Area of Science:
- Sports Science
- Artificial Intelligence
- Biomechanical Analysis
Background:
- Traditional sports training analysis lacks precision for complex athletic movements.
- Shallow machine learning models fail to capture intricate spatial-temporal dynamics of human motion.
Purpose of the Study:
- To introduce a novel deep learning model for precise real-time analysis of athletic movements.
- To enhance posture classification accuracy and robustness in sports training.
Main Methods:
- Developed Deep Dynamic Graph Attention Posture Recognition (DDGAPR), integrating graph neural networks and transformer-based attention.
- Dynamically represented athlete skeletons as graphs, applying spatial and temporal attention to body joint relationships.
Main Results:
- DDGAPR achieved an 18% improvement in motion recognition accuracy.
- Validation accuracy increased by 9.4%, and true positive classes by 9.1%.
- Demonstrated enhanced average attention weight by 10% compared to previous models.
Conclusions:
- DDGAPR significantly improves posture recognition for athletes.
- The model offers potential for personalized performance optimization and enhanced training quality.
- Contributes to AI-assisted sports training and data-driven sports development.