Sistema de reconhecimento facial para registro de consultas clínicas com armazenamento seguro de dados biométricos.

Abstract

This scientific article aims to propose and evaluate a system architecture for clinical consultation check-in via facial recognition, grounded in edge computing and guided by LGPD compliance requirements. The methodology used is based on Design ScienceResearch (DSR), involving the construction of a physical-logical technological artifact and its subsequent computational and mathematical audit. As a theoretical framework, classical works on Deep Learning for facial vector extraction (FaceNet) and current literature on cancelable biometrics and information security in healthcare environments are used. The research is justified by the inefficiency, waiting times, and severe risks of exposing sensitive data inherent in the traditional manual clinical reception process. The results demonstrate that the artifact, operating on a Labrador microcomputer achieves a rate of 20 to 30 frames per second, reducing check-in time to under 1 second passively, even with facial occlusions. The security evaluation proved the effectiveness of data-at-rest protection through AES-128 encryption and biometric irreversibility via an orthogonal matrix (Bio-hashing), successfully resisting reverse engineering simulations. This study contributes to the Health Technology field by presenting a practical and scalable model that solves the administrative bottleneck of patient identification while ensuring maximum privacy by design.