Vision-Based Human Pose Estimation for Intelligent Sports Training and Teaching Assistance

Vision-based human pose estimation is a core technique for intelligent sports training and teaching assistance, enabling fine-grained motion analysis through accurate localization of human body keypoints. However, existing pose estimation methods face significant challenges when applied to real-world sports scenarios. Specifically, fast athlete movements, complex and non-standard postures, severe motion blur, frequent self-occlusion, and dynamic backgrounds lead to substantial degradation in both estimation accuracy and real-time performance. Moreover, many high-accuracy pose estimation models rely on large-scale and computationally expensive networks, which makes them unsuitable for deployment in practical sports teaching systems with limited computational resources and strict real-time requirements. To address these issues, this study focuses on developing a vision-based human pose estimation framework specifically designed for intelligent sports training and teaching assistance. The proposed approach aims to simultaneously improve robustness under complex sports motion conditions and maintain real-time efficiency by adopting lightweight network design and enhanced feature representation strategies. By balancing accuracy and computational cost, the proposed method effectively mitigates the performance degradation caused by motion blur and complex actions. Experimental results demonstrate that the proposed framework achieves competitive pose estimation accuracy with significantly reduced model complexity, making it well-suited for real-world intelligent sports training and teaching applications.