Fingerprint Matching and Non-Matching Analysis for Different Tolerance Rotation Degrees in Commercial Matching Algorithms

Abstract

Fingerprint verification is the most important step in the fingerprint-based biometric systems. The matching score is linked to the chance of identifying a person. Nowadays, two fingerprint matching methods are the most popular: the correlation-based method and the minutiae-based method. In this work, three biometric systems were evaluated: Neurotechnology Verifinger 6.0 Extended, Innovatrics IDKit SDK and Griaule Fingerprint SDK 2007. The evaluation was performed according to the experiments of the Fingerprint Verification Competition (FVC). The influence of the fingerprint rotation degrees on false match rate (FMR) and false non-match rate (FNMR) was evaluated. The results showed that the FMR values increase as rotation degrees increase too, meanwhile, the FNMR values decrease. Experimental results demonstrate that Verifinger SDK shows good performance on false non-match testing, with an FNMR mean of 7%, followed by IDKit SDK (6.71% ~ 13.66%) and Fingerprint SDK (50%). However, Fingerprint SDK demonstrates a better performance on false match testing, with an FMR mean of ~0%, followed by Verifinger SDK (7.62% - 9%) and IDKit SDK (above 28%). As result of the experiments, Verifinger SDK had, in general, the best performance. Subsequently, we calculated the regression functions to predict the behavior of FNMR and FMR for different threshold values with different rotation degrees.