Y. Kim, S. Park, J.-H. Yoo, and K. Choi (Korea)
Bayesian network, Fake detection, Biometric system, Optical flow, Segmentation
In this paper, a novel fake detection approach is presented for biometric face recognition systems using background analysis. More specifically, a probabilistic approach is presented to combine evidences obtained from the background and foreground regions. First, the location of a face is detected by the Adaboost algorithm. Second, the background region, i.e., a region without the face and upper body, is extracted based on the extracted face region. Third, the liveness of current face video can be decided by comparing the similarity of current background region and the original background region that was stored at the initializing stage, and by analyzing the motion of the background region. Lastly, a probability network is used to combine evidences obtained from the background and foreground regions, calculating the probability of the liveness of the current face video.
Important Links:
Go Back
