Facial and Body Gesture Recognition for Determining Student Concentration Level

Abstract

Online learning has gained immense popularity, especially since the COVID-19 pandemic. However, it has also brought its own set of challenges. One of the critical challenges in online learning is the ability to evaluate students' concentration levels during virtual classes. Unlike traditional brick-and-mortar classrooms, teachers do not have the advantage of observing students' body language and facial expressions to determine whether they are paying attention. To address this challenge, this study proposes utilizing facial and body gestures to evaluate students' concentration levels. Common gestures such as yawning, playing with fingers or objects, and looking away from the screen indicate a lack of focus. A dataset containing images of students performing various actions and gestures representing different concentration levels is collected. We propose an enhanced model based on a vision transformer (RViT) to classify the concentration levels. This model incorporates a majority voting feature to maintain real-time prediction accuracy. This feature classifies multiple frames, and the final prediction is based on the majority class. The proposed method yields a promising 92% accuracy while maintaining efficient computational performance. The system provides an unbiased measure for assessing students' concentration levels, which can be useful in educational settings to improve learning outcomes. It enables educators to foster a more engaging and productive virtual classroom environment.