Enhancing eeg-based emotion recognition in Educational environments

Abstract

This thesis addresses EEG-based emotion recognition in educational and smart technologies, highlighting the technical, practical, and ethical challenges of its integration. Emotions play a central role in motivation, engagement, and learning outcomes, yet traditional approaches fail to capture them reliably in real time. To respond to this gap, the research develops and validates EEG-based emotion recognition models using machine learning algorithms applied to both benchmark datasets and data collected with consumer-grade devices. The methodology emphasizes robust preprocessing and classification pipelines that maintain accuracy under noisy and diverse conditions, with a particular focus on reducing the electrode set to key frontal channels (Fp1, Fp2, F3, F4) to simplify data collection and retention without compromising performance. The findings demonstrate that robust models can be achieved with simplified EEG configurations, offering practical feasibility for educational contexts and greater adherence to privacy requirements. Although classroom deployment was beyond the scope of this work, the thesis lays the foundation for future research on federated learning, attention modeling, and multimodal affect detection, advancing the responsible use of affective computing in education.