Employing the OpenPose model, this research advances the prediction and analysis of human skeletal points in videos, paving the way for refined action differentiation and posture analysis. By isolating unique posture features and developing a DTW-based algorithm for posture matching, we identify deviations in movements for corrective action in sports training. This methodology promotes the digitization of physical education, enabling personalized coaching and a detailed understanding of student performance. Our findings reveal a marked improvement in the experimental group’s movement evaluation scores, surpassing the control group by 82.86 points, with a P value of 0.000, and demonstrating significant skill test advancements (t=3.129, P<0.01). These results validate the approach’s potential to revolutionize physical education management and sports training with data-driven insights.