Estimation of Student’s Engagement with a Pressure Mat and a Web Camera

Abstract

It is important for teachers to grasp students’ engagement in order to improve the quality of lectures. When they find that their students are not engaged in the lecture, they can give advice to the students to pay attention to the lecture. However, in the e-learning environment, there is no teacher to grasp the student’s engagement. So if the student loses his engagement, no one can help him to regain it. It may cause ineffective learning. The purpose of this study is to grasp the student’s engagement by using a pressure mat and web camera. We recorded the students’ postural data, that is upper body pressure distribution and upper body pose information, while they were taking e-learning lectures. In order to get the body to pose information from a web camera, we used a human pose estimation library called OpenPose. Then we extracted 38 features from upper body pressure distribution and 33 features from upper body pose information for every minute. We selected effective features by using the Forward Stepwise Selection. Lastly, we estimate whether he or she was engaged in or not with a Support Vector Machine. As a result, the average accuracy was 79.3% for student-dependent estimation. This result shows it is possible to predictthe student’s engagement automatically.