Implementation of Real-Time Facial Emotion Recognition using Advanced Deep Learning Models

Facial emotion recognition (FER) is a crucial technology in human-computer interaction, enabling machines to understand and respond to human emotions effectively. This research aims to develop and implement a real-time FER system that accurately classifies facial expressions using state-of-the-art deep learning models. Specifically, we implement and compare four advanced FER models: EmoNeXt, Deep Automatic Facial Expression Recognition Model (DAFERM), Dual IntegratedCNN (DICNN), and LibreFace. The system processes live video input, detects faces using OpenCV, and classifies emotions into predefined categories using deep neural networks. The models are trained on benchmark FER datasets such as FER2013, CK+, and AffectNet. Real-time deployment is achieved using a Flaskbased web application for interactive visualization. Performance is evaluated based on classification accuracy, inference speed, and latency across different hardware configurations. The results demonstrate that DICNN achieves the highest accuracy (93.2%) with a balanced inference speed, while LibreFace offers the fastest real-time performance. The proposed implementation has potential applications in AIdriven chatbots, mental health monitoring, and interactive gaming. Future work will explore multimodal emotion recognition, integrating voice and gesture analysis for enhanced interaction.