In order to improve the teaching effectiveness and training personalization level of frisbee sports classes in private universities, this study proposes the use of automated detection technology for automatic measurement of frisbee sports training process. Firstly, the overall structure of the system covering the functions of physiological information detection and movement data detection was designed, and then the construction of the dataset of the AI teaching assistant application system for physical education classes was completed, from which the human skeletal key point data were extracted. Then the spatial and temporal relationships between joints were modeled using spatio-temporal graph convolutional networks. Finally, the skeletal coordinates are converted into joint angles, and the DTW distance is used as a parameter to define the action evaluation formula to evaluate the Frisbee sports action. Experiments show that the disc sports movement recognition accuracy of the AI teaching assistant system for disc sports classes in private colleges is more than 90%, and the error between the algorithmic score and the expert score is only 0.7 points, which can effectively correlate the relationship between the disc sports movement and the movement extension angle, which verifies the feasibility of the teaching assistant system in the auxiliary teaching of disc sports classes in private colleges and universities, and it can realize the accurate and personalized training of the students.
