"Emotion Detection recognition using Machine Learning Intelligence "
Year: 2026
Type: dataset
Abstract: "EmoFusion is a multimodal deep-learning framework designed for robust real-time emotion recognition by combining facial expressions and speech. It uses EfficientNet-B0 for facial feature extraction, SUPERB-ER for Transformer-based speech representations, and a dynamic weighted late-fusion engine for modality-level predictions. The framework was evaluated on RAVDESS, FER2013, CREMA-D, and CMU-MOSEI using random-split evaluation, with additional speaker-independent LOSO validation on RAVDESS. It achieved accuracies of 87.11% on RAVDESS, 89.21% on CREMA-D, and 84.22% on CMU-MOSEI, while achieving 83.26% under LOSO evaluation. The system supports real-time deployment and is designed for applications in healthcare monitoring, human-computer interaction, behavioral analytics, and emotion-aware intelligent systems. Its multimodal architecture improves robustness against variations in individual audio or visual inputs."
Source: IEEE DataPort
DOI: 10.21227/r5qb-5y12
Authors: Sanchit Borikar, Veer Gandhi, Harshit Jain
Language: English
Topic: Emotion and Mood Recognition
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204436096
