Eyes Sequence Analyzing Based Face Liveness Detection Method

With the development of the biometric technology, biometric systems have been applied in almost all aspects of human life. Face recognition has become one of the most widely used biometric technologies, because it is user-friendly and non-intrusive. However, current face recognition systems are vulnerable to spoofing attacks. Attackers can treat the system using a face photo of any authorized user. The face liveness detection technology determines whether the face image sequence sampled from the system is from a living person. This is an effective means of improving the safety of face recognition system.Researches of face liveness detection can be roughly divided into two categories. One category deals with the materials used to spoof the system. According to the optical flow or texture produced by three dimension al pictures and two dimensional objects, they can distinguish living faces and faces in pictures. Another category exploits the features of a living body. They detect liveness by identifying the physiological movement of some special areas of a living face.This research focuses on the eyes area that presents the most notable living features on the face region. This research performs face liveness detection via the analysis of sequences of living eyes and that of eyes in pictures.A face liveness detection method based on hidden Markov is proposed. This method trains the hidden Markov models of image sequence of living eyes and that of eyes in pictures. Then it can determine whether the eye image sequence being detected is from a living person or not, based on to the likelihood degree of the detected sequence with different models.A face liveness detection method based on the low rank analysis of an eye image sequence is designed. Due to the fact that the image sequence of living eyes changes more than that of eyes in pictures, this method is able to distinguish a living eye image sequence from image sequence of eyes in pictures by analyzing their image matrices.The main contribution of this research is putting forward a new classification basis for liveness detection. By setting up a model and completing a series of theoretical analysis, this research elaborates how the basis is produced. Experimental results on relative data sets show that the classification basis can be used in face liveness detection effectively.