Research on Artificial Intelligence Application Strategies in the Integration of Physical Education Teaching and Athletic Training for College Students in Higher Education

Abstract

In this paper, the hourglass network is used to extract the key points of human skeleton of students during physical education teaching and sports training, and to estimate the students' single movement posture characteristics by eliminating the abnormal frames and coordinate transformations, parameter calculations, and division of possible movement intervals. Based on the differences in the movement posture features, the DTW algorithm is used to calculate the similarity between the student's movements and the teacher's movements during physical education and sports training, and score them accordingly. In the key action posture matching, the matching success rate of this paper's method for six types of actions is between 70.52% and 90.43%, which is higher than that of other methods. The classification accuracy calculated by this paper's method based on similarity is high, and the number of classification errors for the 6 classes of actions is mostly in the range of 0-3 times. The number of model parameters of this paper's method on 2 validation sets is only 13.1M, which is lightweight. In the motion pose matching deviation rate experiments, the scoring deviation rate of this paper's method for the excellent and unqualified phases ranges from 3.16% to 5.56%, and the overall scoring deviation is in line with expectations.