Gabor feature extraction of mixed fingerprint template

Abstract

Biometrics based authentication systems face the major challenge in securing the template. Although multi-modal biometric techniques and cryptographic techniques offer security but they can be easily compromised. This paper explores the possibility of combining features from the same biometric trait (fingerprints) i.e. keeping it unimodal but still offering the same security as that of the combined biometrics. In this paper, using minutiae and orientation features from two different fingerprints, a new unique mixed fingerprint template is generated. The advantage of creating the mixed template is that it is cancellable and serves as a new virtual identity. Further Gabor feature of the mixed template is extracted before matching to reduce the error rate. Testing the templates is done using 2 methods, viz: correlation based and minutiae based. These two methods were effective in identifying the templates accurately. The FAR=0.08% and FRR=0.1%.