Abstract:

The coronavirus disease (COVID-19) has created a need for more extensive distance education. This has led to a distinctive rise in e-learning. A learning management system (LMS) is a software application that supports e-learning. Teachers can use an LMS to efficiently manage the student learning processes. However, there are some classes in which data management and automatic scoring by LMS are difficult. Physical education is one such class. Although a teacher can observe a student's physical performance via a remote video camera, and record it as digital data, the evaluation is done visually. Therefore, data analysis and data management cannot be automated. Furthermore, the teacher's visual acuity is heavily burdened when multiple videos are being visually evaluated on a computer display. To address this problem, we have developed an automatic evaluation system for students’ exercise demonstration videos. The proposed system detects the human skeleton from the video and quantifies joint movements to automatically evaluate the performance of the exercise. The skeleton detector was developed based on a deep neural network (DNN) framework. Experiments were carried out using a video of a basic exercise to evaluate the system.

Keywords:

e-learning, physical education, automatic evaluation system, deep neural network.