Facial action units for presentation attack detection
Pan Shi, Farzin Deravi · 2017
This paper is concerned with biometrie spoofing detection using the dynamics of natural facial movements as a feature. Facial muscle movement information can be extracted from video sequences and encoded using the Facial Action Coding System (FACS). The proposed feature constructs a Facial Action Units Histogram (FAUH) to encapsulate this information for the detection of biometric presentation attacks without the need for active user cooperation. The performance of the proposed system was tested on two datasets: CASIA-FASD and Replay Attack and produced encouraging results. Further improvements may be possible by integrating this source of information with other indicators for further protecting biometric systems from subversion.