Fingerprint Liveness Detection with Feature Level Fusion Techniques using SVM and Deep Neural Network

Fingerprint liveness detection has become a common technique in fingerprint authentication systems for security purposes. Although several hardware and software based approaches have been introduced so far to distinguish between live and fake fingerprints, finding effective features for detecting the fingerprint liveness still remain unsolved. In this paper, a liveness detection method is proposed, which combines discriminative features obtained from Speeded-Up Robust Features (SURF), Pyramid Histogram of Oriented Gradients (PHOG) and texture features from Gabor wavelets. Apart from using these feature extraction methods individually, this paper mainly focus on feature level fusion of suggested methods. Experiment is done with Support Vector Machine (SVM) and Deep Neural Network (DNN) classifiers over LivDet 2013 database. It is found that the classification accuracy of the proposed detection technique is higher for DNN as compared to SVM classifier for different feature level fusion techniques.