A Vision-Based Behavioral Measurement Using Facial Expression for Detecting Student Concentration in Classrooms

Effective teaching requires the facilitator to gauge the students' concentration levels during a lecture and accordingly elaborate on a particular topic or repeat a concept that students did not understand. The current methodology for in-classroom monitoring of concentration levels is by asking the students whether they have understood, or conducting a postassessment quiz. We propose a minimally intrusive computer vision-based approach to obtain facial expressions for behavioral measurements. Deep Learning models are used to detect emotions. A pairwise comparison and rank reciprocal is used for assigning weights for various emotions corresponding to behavioral engagement in the class. A sampling method is proposed to reduce computation time and calculate the aggregate concentration index over a time horizon. The results show that the facial expression of students during a lecture is a reasonably accurate metric in detecting the concentration levels of a class and may be implemented to provide real-time feedback.