Fingerprint templates generated by the fractional fourier, cosine and sine transforms and their generation conditions

Abstract

Abstract�Recently, we have proposed new fingerprint templates based on the ID discrete fractional Fourier, cosine and sine transforms (DFRT, DFCT and DFST) to realize the fingerprint recognition system with high recognition accuracy and high robustness against attacks. In the previous study, I evaluated the recognition accuracy and robustness of the generated templates of the fingerprint data of the FVC 2002. As a result, I found the most appropriate templates are the amplitude distributions of the DFCT and DFST with a size of 96 by 96 pixels under the condition that the range of the transforms' orders is between 0.1 and 1.0. The EER was an order of 10. 6 % and the PSNR was an order of several dB. However, a set of the transforms' orders was changed for different fingers. In this study, the appropriate templates are decided under the condition that only one set is used for different fingers. As a result, the most appropriate templates can be obtained under the same generation conditions, though only the range of transforms' orders is different, i.e., between 0.3 and 1.0. The EER is 1.86% less than that (2.65%) of the original fingerprint images. The PSNR is an order of several dB.