Biometric system to prevent unauthorized access through facial and emotional recognition in critical environments

De Jesus Lopez Coral T.; Ronals Lopez Acuna F.; Huarcaya Junes A.N. · Proceedings - 2025 International Conference on Cyber-Enabled Distributed Computing and Knowledge Discovery, CyberC 2025 (2025) · DOI: 10.1109/cyberc66434.2025.00019

Resumen

This article describes the design, development, and implementation of a biometric access control system that integrates facial recognition with automatic emotion detection, applied to critical healthcare facilities. This solution addresses the shortcomings of traditional card-based methods, which are vulnerable to impersonation, theft, or cloning attacks. The system authenticates the identity of authorized personnel using deep learning techniques and simultaneously assesses their emotional state based on facial expressions, using a convolutional neural network trained on the FER-2013 set. When emotions considered critical (such as fear, stress, or anger) are detected, a silent alert is generated without blocking access, prioritizing timely human intervention. The prototype was validated in a controlled environment, achieving an accuracy of 92.4% in facial recognition and 86.3% in emotion detection, with average response times under 3 seconds. The system's modular architecture, based on microservices on.NET Core 8 and ONNX models, allows for its integration into real-world environments and its scalability to multiple access points. The results obtained demonstrate the effectiveness of the proposed approach for improving proactive security in sensitive areas, combining artificial intelligence techniques with usability principles and respect for biometric privacy.

Instituciones: Universidad Peruana de Ciencias Aplicadas