Research On Texture Feature Based Fingerprint Liveness Detection

With the development of information technology, identity authentication plays an increasingly important role in information security field. By relying on the unique biological characteristics of each individual, biometric systems are often used for authentication in various security applications, and the traditional identity authentications are gradually replaced by biometrics. Fingerprint has the highest occupation rate of the authentication mode because of its uniqueness and easiness to collect. However, the fingerprint recognition systems are vulnerable to spoof attacks as fake fingers can be easily produced with various materials. Thus, it is important to equip the fingerprint recognition systems of the spoof fingerprint detection ability.Based on the texture features of fingerprint images, this paper presents two fingerprint liveness detection methods to detect the spoof fingerprints. Summary, the specific research contents of this paper are summarized as follows:1) Fingerprint liveness detection using gradient co-occurrence matrixAccording to the characteristics of fingerprint image,this method extracts texture features from image gradients. The quantization and truncation operation is conducted on the gradients to reduce its range and the dimension of feature. The proposed gradient co-occurrence matrix (GCM) is constructed from the gradients of the adjacent pixels, and effectively reflects the texture difference between the live and spoof fingerprints. The support vector machine (SVM) is used as classifier and the performance of the algorithm is evaluated on the standard fingerprint image database.2) Fingerprint liveness detection based on Weber local binary descriptorBased on the deficiencies of Weber local descriptor (WLD) on texture description, this paper presents the Weber local binary descriptor (WLBD) by combining the original WLD and LBP. It consists of two components: local binary differential excitation (LBDE) extracting local contrast features by combining local binary pattern and Weber's law, local binary gradient orientation (LBGO) extracting gradient orientation features from centrosymmetric pixels. The co-occurrence probability of the two components is calculated to construct a discriminative feature vector and the effectiveness of the algorithm is evaluated by algorithm analysis and experiments.At last, the presented work was summarized and the further research of fingerprint liveness detection was discussed.