Develop, Design, and Evaluate a Real-Time Facial Micro-Expression Recognition Model for Mental Health Monitoring Using a Three-Dimensional Convolutional Neural Network

Facial Micro-Expression (FME) is a subtle, involuntary movements that reveal genuine emotions, making them critical indicators in psychological evaluation. This paper presents a lightweight, real-time-capable facial Micro-Expression Recognition (MER) framework using a 3D CNN architecture for passive mental health monitoring in telehealth settings. The model leverages an R(2+1)D convolutional structure to capture spatio-temporal patterns in video clips and achieves high performance on the CASMEII and SAMM datasets, with F1-scores of 83% and 80%, respectively. It maintains low inference latency (≤ 35ms) and meets real-time speed requirements (≥ 28 FPS), positioning it as a practical solution for affective computing applications. Comparative analysis with state-of-the-art methods demonstrates its balance between accuracy and deployability.