Enhancing Masked Face Recognition with Real-Time Eye Blink Detection for Secure Access Control

This paper presents a novel approach for enhancing masked face recognition in access control systems through the implementation of eye blink detection, utilizing 68 facial landmarks and the eye aspect ratio (EAR). It examines the role of these markers and the use of eye landmarks to accurately calculate EAR. Access systems, especially those used in banking applications, often rely on passwords or multi-factor authentication methods such as passwords combined with facial recognition. However, these traditional methods have certain vulnerabilities, including susceptibility to shoulder surfing and facial spoofing. To address these challenges, an improved masked face recognition method that integrates eye blink detection is proposed, providing a robust solution for access control and liveness detection. The masked face recognition component employs a deep learning model, achieving a 99.77% recognition accuracy on a benchmark dataset, with real-time eye blink detection built onto this model to prevent spoofing attacks via static images. To showcase the viability of this approach, a web application called "MaskBlink" has been developed, with functional tests conducted to validate its key features.