Deep learning for finger-knuckle-print identification system based on PCANet and SVM classifier

Abstract

Biometric technology knows a large attention in the recent years. In the biometric security systems, the personal identity recognition depends on their behavioral, biological or physical characteristics. Currently, a number of biometrics technologies are developed and one of the most popular biometric trait is finger-knuckle-print (FKP) due to the user-friendly and the low cost. This paper presents a new approach, where the deep learning is applied to create a multi-modal biometric system based on images of FKP modalities which extracted their features by principal component analysis Network (PCANet). In the proposed structure, PCA is employed to learn two-stage of filter banks followed by simple binary hashing and block histograms for clustering at feature vectors, which is adopt as input for classification by linear multiclass Support Vector Machine (SVM). To improve the recognition rates, a multimodal biometric system based on matching score level fusion scheme was generated. Using an available FKP database, we conducted a series of identification experiments and the obtained results show that the design of our identification system achieves an excellent recognition rate and having high anti-counterfeiting capability.