Improving facial recognition accuracy by applying liveness monitoring technique

摘要

In a typical transmission control protocol and Internet protocol (TCP/IP) suite, there are several ways, such as using photograph, three-dimensional (3D) model, and video clip of a valid user to mock the facial recognition. Studies show that there has been significant improvement in detecting photograph and 3D model spoofing. However, there is no such improvement in detecting video spoofing. Recent studies suggest that liveness monitoring using the facial features has potential to improve security, especially in biometric systems. In this paper, a liveness monitoring technique is introduced to assist facial recognition. In the proposed system, an eye blinking pattern is used through a real-time generic web-camera. An eye blinking password system using ultrasonic range sensing module is developed for liveness detection of the users. The system is tested by conducting experiments using 15 valid users and 100 different user appearances. According to the experimental results, the proposed system achieves 99% face recognition accuracy and 99% liveliness detection performance.