People can have access to biometric system easily by face spoofing attack. Recent researches have proposed many anti-spoofing strategies based on eye blinking, facial expression changes, mouth movements or skin texture. As for the face which has slight trembling, there are few specific methods about it. To solve this problem, we have proposed a method which could discriminate human real face and face print by using parameters of slight face motion and equal-proportion property in projection. In order to validate our method, we also established a new video database containing 720 moving faces. After getting the facial landmarks, aligning image of each frame and calculating the variance as feature value, final data would be sent to the support vector machine (SVM) to verify the reality of faces. From the experiment results, the proposed method shows its high accuracy in different lighting conditions and different face amplitude for anti-spoofing and will have good prospect in engineer.
展开▼
