Intelligent Recognition of Physical Education Teachers' Behaviors Using Kinect Sensors and Machine Learning

In this research, Kinect sensors were used to obtain body posture data of physical education (PE) teachers during simulated classes and in combination with classical algorithms of machine learning, to achieve the intelligent recognition of the classroom teaching behaviors of PE teachers. Kinect 1.0 was used to test 10 PE teachers without students during simulated classes, and the characteristics of body postures corresponding to different teaching behaviors during the classes of PE teachers were obtained through time sampling. The accuracy of the light gradient boosting machine (LightGBM) recognition model combined with the Kinect sensor was 0.998, which was significantly higher than those of other algorithms. The combination of Kinect sensors and machine learning enabled the intelligent classification of, for example, password teaching, language explanation, action demonstration, and guiding behavior during a simulated class of PE teachers. The recognition models trained by LightGBM were the most effective.