Soft Computing Approaches to Emotion Recognition in Virtual Learning Environments

Abstract

Emotion recognition in virtual learning environments (VLE) is critical for adaptive tutoring and personalization. This study investigates hybrid soft computing frameworks combining fuzzy logic, neural networks, and genetic feature selection to infer learner affect from facial expression, voice, and interaction patterns. Trained and evaluated on 250 learners in a VLE context, the hybrid model achieved 87.5% accuracy (±2.1), outperforming standard CNN (82.3%) and fuzzy-only systems (79.4%) in recognizing four basic affective states. Early detection latency was reduced by 15%, enabling real-time affect-aware interventions. These findings support the integration of soft computing for responsive and emotionally intelligent VLEs.