Smart Education Platform: Based on Emotions Content Delivery Using ML

Abstract

This project proposes a Smart Education Platform that dynamically adapts online learning content based on students’ real-time emotions to improve engagement and outcomes. Leveraging multimodal inputs—facial expressions detected via webcam using ResNet, speech sentiment analysis through Wav2Vec, and contextual surveys the system identifies states such as engagement, boredom, and annoyance. Integrated within a Flask based web application and Moodle, a reinforcement learning agent tailors interventions (quizzes, lessons, condensed material) by optimizing a reward function (+1 for engagement,-1 for negative emotions). Privacy is prioritized through webcam opt-out, on-device processing, and informed consent. The solution was validated on 20–50 learners using AffectNet and other datasets, achieving high accuracy with fivefold cross-validationThe system provide personal, scalable, and privacy-aware experiences in a variety of educational contexts, addressing the limitations for static learning platforms. A focus on emotional agility enhances efficiency and motivation, setting a standard for future massive deployment and human-centered digital learning.