Face Liveness Detection Based On Spectrum Analysis And Depth Information

With the maturing of face recognition technology,its commercial products are more and more widely used in our life.However,with more implementation of that technique,some serious safe issues occur.Currently,almost all face recognition systems cannot distinguish the real from fake face.In other words,they cannot judge whether the face ahead of the camera is fake or real human face and they tend to be cheated by the 3D masks,videos,or photos of the human face.The face liveness detection aims to solve the potential above-mentioned risks in face recognition technique.The face liveness detection provides the system with the ability for identifying the real from fake face and it can ensure the safe and stable operation of the face recognition system.The research in this paper mainly contains the following aspects:(1)This paper comes up with the method of face liveness detection based on the hyper-spectral analysis for the first time,uses the hyper-spectral imager based on the liquid crystal tunable filters to collect the hyper-spectral data of human faces,and analyzes the spectral features of the real human faces and the photos of human faces under various light conditions and various feature positions.Meanwhile,it is compared with the method of face liveness detection based on the multi-spectral analysis through experiments.(2)This paper comes up with the method of face liveness detection based on the depth information,uses Intel RealSense 3D camera to collect the 3D depth data of the real human faces and the photos of human faces,and analyze the 3D features of the real and fake human faces.It mainly studies the 3D geometrical features,expression change and gesture change on human faces.(3)This paper builds a data base of face liveness detection which contains 20 real human faces and 80 photos of human faces and verifies the method proposed in this paper based on the data base through experiments.The experiments results show that the method proposed in this paper can solve the problem of face liveness detection efficiently.