A Research Review of Intelligent System for Recognition Human Emotion in Real-Time Using AI

Human Emotion Recognition (HER) is a cornerstone of the burgeoning field of Affective Computing, facilitating a more intuitive and symbiotic relationship between humans and machines. This research provides an in-depth analysis and proposal for an intelligent system capable of real-time facial emotion recognition (FER) utilizing advanced Artificial Intelligence (AI). While historical methodologies focused on static image analysis under controlled conditions, contemporary requirements demand high-performance, real-time video stream processing in unconstrained environments. This paper details a comprehensive software architecture that integrates Multi-task Cascaded Convolutional Networks (MTCNN) for face detection and lightweight Convolutional Neural Networks (CNN), specifically the Mini-Xception architecture, for emotion classification. The study addresses the critical "trilemma" of modern AI: maintaining high accuracy, achieving sub-100ms latency, and ensuring robustness against real-world digital noise such as varying illumination and facial occlusions. Through an extensive literature review, the paper identifies systemic gaps in micro-expression detection and cross-cultural generalization. The proposed methodology emphasizes a modular, multi-threaded approach to ensure a seamless Graphical User Interface (GUI) experience. Expected outcomes include a validated system capable of maintaining ≥ 30 FPS on standard hardware with an accuracy exceeding 92% on benchmark datasets, paving the way for applications in telehealth, automated driver monitoring, and adaptive pedagogical tools.