Intelligent Monitoring of Physical Education Classroom Teaching in Higher Vocational Education Based on Hybrid Deep Learning Model

In Physical Education (PE), students struggle with precise movements throughout their early motor skill learning. Correcting and eliminating these inaccurate movements is crucial for teachers to enhance the quality of PE. This research develops an Intelligent Monitoring System (IMS) for PE classroom teaching in higher vocational education based on Deep Learning (DL). In this work, the method of combining MobileNetV3, Single-Shot-Multibox Detector (SSD) and Bi-directional Feature Pyramid Network (Bi-FPN) is used for effective motion detection and feedback. The basis of SSD is MobileNetV3 which is designed for mobile vision tasks, while the Bi-FPN enhances the fusion of features at different scales to enhance motion detection. The system ensures to provide real-time analysis and correction to make a feedback to the students. The overall research provided enhanced performance for mobile application, which leads to the better participation of students and better fitness values. Experimental results showed that the proposed hybrid DL model achieved better accuracy at 99.27%, precision at 97.73%, recall at 98.64%, and error rate at 1.26 when compared to the existing methods such as Convolutional Neural Network (CNN) regression, and Deep Neural Network (DNN).