The rapid development of artificial intelligence in the field of education has injected new momentum into the systematic transformation of educational models, driving innovation and optimization in teaching methods, evaluation systems, and management models. This paper uses the OpenPose algorithm to obtain skeletal point data for sports movements and standardizes the skeletal coordinate data. Based on the ST-GCN model, a multiscale temporal attention mechanism is introduced to enhance the model’s feature extraction capabilities, and a residual module is incorporated into the GCN to improve the model’s local feature extraction performance. On this basis, the DTW algorithm is used to construct a sports movement evaluation model. Experiments show that the improved ST-GCN model achieves a MAP value of 79.3%, which is 9% and 6.8% higher than the MAP values of the image-based pose estimation algorithms SimpleBaseline and HRNet, respectively. The overall score for sports actions based on the DTW algorithm is 88.19 points, differing from the manually scored results by only 1.12 points. Integrating artificial intelligence technology with sports education and teaching can significantly enhance the intelligent reform of sports education and provide a technological foundation for the personalized development of sports education.
