Abstract

Emotional awareness is crucial for effective learning, yet students’ emotional cues are often overlooked
in large classrooms. This study presents a fully offline facial emotion recognition system that assists
educators in monitoring student engagement and emotional states without relying on external databases
or internet connections. Leveraging a Convolutional Neural Network, the system classifies seven
universal emotions: happiness, sadness, fear, anger, surprise, disgust, and neutrality. The system
achieved high predictive performance, with an overall accuracy of 92%, precision of 92.3%, recall of
91.3%, and F1-score of 91.8%, demonstrating reliable real-time emotion recognition. By providing
timely insights into students’ emotional responses, the system enables educators to adapt teaching
strategies, enhancing both academic engagement and emotional well-being. While performance may
vary with subtle emotions or cultural differences, the approach offers a practical tool for creating
emotionally responsive learning environments in classrooms with limited technological resources.

Key words: Facial Emotion Recognition; Deep Learning; Convolutional Neural Network (CNN); Real time Emotion Detection, Offline deployment