Aiming at printing attacks and video replay attacks in face recognition,this paper proposes a face anti-spoofing algorithm based on tiny motion.We perform motion amplification algorithm to the video to make tiny motion more obvious.Two motion feature maps were created and fused by using motion intensity and motion direction to describe the tiny motion.A VGG16 network based on attention mechanism was adopted to distinguish the real faces and fake faces.The half total error rate(HTER)of our algorithm on the Replay-attack dataset is 1.35％and the equal error rate(EER)on the CASIA FASD dataset is 2.5％,which proves that the micro motion is effective to prevent facial counter-feiting.
Face anti-spoofingMotion amplificationFacial movementsMotion feature mapVGG16
国家科技期刊平台
NSTL
万方数据
