Evaluation of online teaching quality based on facial expression recognition

Abstract

With the rising prevalence of online lessons, it has become increasingly clear that teachers are unable to determine the students’ engagement levels through the screen. As such, this paper proposes a system to a deep learning model for facial expression recognition to determine the engagement of students. This is done by analysing the students’ facial expressions to classify their emotions throughout the online lesson. The facial expression recognition information is used to calculate the engagement index, which is then categorised into one of four engagement states: “Highly Disengaged”, “Disengaged”, “Engaged”, and “Highly Engaged”. Publicly available datasets CK+, RAF-DB, and FER2013 are used to gauge the overall performance and accuracy of the proposed model. Experimental results showed that the proposed model achieves an accuracy of 0.9697, 0.8625, and 0.8451 on the datasets CK+, RAF-DB, and FER2013 respectively. On the dataset combining all the earlier mentioned datasets, the proposed model also achieved an accuracy of 0.8561. A frontend application is also created and linked to the trained model to record the video of students during the online lesson and evaluate their engagement states. This system provides teachers with useful information to enhance student engagement in future lessons.