Improvement of Fingerprint Verification by Using the Similarity Distribution

Mobile devices, with their excellent portability and increasing computational power, are increasingly being used for communication and financial transactions. As they are used in close relation to people, their security is becoming more important. Faceless verification systems with improved security performance, including face or fingerprint verification, are recently being required. Fingerprint verification is a suitable method in a faceless environment. However, the commonly used Minutiae-based fingerprint verification shows a drop in the performance of fingerprint verification, due to the decreased number of minutiae, when the number of acquired images is small. Especially since the values around the threshold of similarity are similar in the genuine and imposter, many errors could occur here. The minutiae-based fingerprint verification has a limitation in addressing these problems. A hybrid-based verification method that uses two or more fingerprint matching methods can address these problems better. Therefore, this paper has conducted the binary-image-based fingerprint verification in the partial band around the threshold. From the results of the experiment, it can be seen that the Equal Error Rate (EER) was improved by a total of 42 %, from 3.01 % to 1.73 %, by reducing the False Match Rate (FMR) in the partial band area around the threshold. In addition, it was improved by reducing the FMR by a total of 89 % from 2.77 % to 0.28 %.