FaSIVA: Facial signature for identification, verification and authentication of persons

Abstract

The artificial reproduction of human vision capabilities has become unavoidable in recent technological applications, and as the number of applications increases, more challenges demanding specific attention in the proposition of solutions arise as well. Among this wide range of applications, there is face recognition in general and in video surveillance and access control where a lot of models have been proposed to suit users’ needs. Although lot of progress has been made by proposing models for face recognition, but still there are issues affecting recognition accuracy. These issues include low resolution of input image and effective presence before capture equipment’s. To address these issues, this paper presents a formal representation of face called face signature that considers three parameters namely, the quality of the input image, the patterns extracted from the image, and the characteristics of human presence namely eye blinking and liveness detection. Our experiment results show that the proposed signature is accurate and suits in real life applications where input images are of low resolution and spoofing attack are used. • Enhancing the overall face recognition process in terms of recognition accuracy. • Prevention of spoofing attacks can be possible using the detection of eye blinks. • Optimizing deep CNN architectures for face liveness detection. • Face Recognition (FR) is useful for solving access control issues.