Facial Recognition Using Edge-Driven Biometric System
Chandrakala G. Raju, Beena Ullala Matha, Varun Canamedi, Somdyuti Sarkar, K R Radhika · 2022 International Conference on Applied Artificial Intelligence and Computing (ICAAIC) · 2022
Non-invasive biometric methods such as facial recognition reduce the risks and difficulty that come with handling confidential biometric data. Also, it makes the task of providing security simpler while ensuring accurate results simultaneously. Integrating biometric systems with Edge Computing and Deep Learning, makes the system more robust and dynamic by reducing latency and bandwidth usage. The purpose of this paper is to present an optimal facial recognition model suitable for a wide range of applications. The system uses HOG descriptors combined with Deep Learning to identify a person from a custom database. The facial recognition system is hosted on low-power devices with the help of an Internet-of-Things (IoT) network, making it entirely edge-based. A standard Message Queue Telemetry Transport (MQTT) protocol hosted over a local Wi-Fi network is used to facilitate communication between devices in the network. An accuracy of 98.33% has been achieved.