Totem de reconhecimento facial em computação de borda: construção de um artefato tecnológico sob a ótica da Design Science Research

Abstract

Manual patient credentialing in healthcare institutions generates exhausting queues, consumes excessive time, and exposes the environment to security and hygiene vulnerabilities. To overcome these operational limitations, this scientific article aims to develop a facial recognition technological artifact based on Edge Computing architecture to automate the patient credentialing process, reducing wait times and mitigating operational failures. The methodology used is based on Design Science Research (DSR), with a quantitative approach and experimental field research. The system was built by integrating a Single Board Computer (SBC Labrador) with an ESP32-CAM module and distance sensors (VL53L0X), operating entirely offline and contactless. The computer vision pipeline utilizes the OpenCV and Dlib libraries in conjunction with Deep Neural Networks for extracting mathematical facial signatures (embeddings). The evaluation phase cross-referenced the empirical baseline of the manual process in clinics — which recorded wait peaks of up to 7 minutes and 51 seconds per patient — with the prototype’s performance in a controlled environment. The results demonstrate that the artifact completes the inference and identity validation cycle in approximately 1.3 seconds. It is concluded that the proposed solution successfully meets the outlined objective, converting a time-consuming in-person procedure into an instantaneous validation, structurally eliminating wait queues and reducing human error in the clinical environment.