Evaluating Students’ Reaction to Lectures Using Facial Expression Recognition

Abstract

Interaction between lecturers and students plays a critical role in defining one's view of a lecture. However, with an increasing number of students enrolled in universities every year and limited classroom space available, classrooms are often overcrowded. As a result, it is fairly difficult for lecturers to immediately observe the learning feedback from all students on the lecture being delivered. In this paper, we propose a tool named Facial Expression Analysis Tool (FEAT) to help lecturers in universities in evaluating the effectiveness of their lecture by evaluating their students’ facial expression based on three facial expressions: bored, satisfied or confused. The tool utilizes dual CNN for detection and classification. FEATreceives the video feed via an IP camera from the classroom, and analyzes and stores the information on a cloud database. The aggregated information from the database is further filtered, and the statistical details are displayed on a visual dashboard on the web. The tool was evaluated in a real classroom environment and found to have achieved a good accuracy. The tool provides useful insights for the lecturers to better observe their students’perception on their lectures and improve their teaching approach if required.