Real-Time Emotion Detection in Online Learning Environments Using CNN-Based Facial Expression Analysis and Alert Mechanism

Abstract

Online education has become increasingly prevalent, but it often lacks the emotional awareness present in physical classrooms. In virtual settings, teachers are unable to gauge students’ emotional states - such as boredom, confusion, frustration, or engagement - which play a critical role in the learning process. To address this problem, our project proposes a realtime emotion recognition system that monitors student emotions during online classes through live webcam analysis. The system uses a Convolutional Neural Network (CNN) to detect and classify facial expressions into distinct emotional states. Real-time video is captured using OpenCV, and facial features are extracted from each frame for emotion prediction. Publicly available datasets such as FER-2013 (for basic emotions) is used to train the model, ensuring reliable recognition of student states in educational scenarios. To enhance the learning environment, the system integrates a feedback alert mechanism that notifies mentors when students display signs of negative emotional states like prolonged confusion or frustration. Additionally, the project features an emotion timeline visualization, which plots variations in student emotions throughout the class session, providing mentors with valuable post-session insights. This intelligent and scalable system brings emotional awareness into digital classrooms, enabling more responsive and empathetic teaching. Designed to be achievable within a student project scope, it combines computer vision, deep learning, and human-centered design to improve virtual learning outcomes.