Recognition Accuracy and Robustness of Fingerprint Templates Generated by Discrete Fractional Sine Transform

Abstract

The fingerprint template which corresponds to the phase distributions (PDs) Individual recognition using the biological information has been recently increasing everywhere, for example, in the automatic logging into a PC, the immigration at the airport, and so on. In particular, the fingerprint recognition system has been widely used because of its high reliability and reasonable price [1].In general, the enrolled fingerprint images are stored as templates in the database referred in the fingerprint recognition process. , and so on. However, in these methods, since the minutia-based templates are basically used, the recognition accuracy tends to become low in comparison with that of the image-based templates.In order to realize the fingerprint templates with high recognition accuracy and high robustness against attacks, we have proposed a new method for generating templates by use of the intensity distributions of the 1D fractional Fourier transforms (FRTs) [11] of original fingerprint images [12,13]. In our method, the FRTs with different transforms' orders are applied to the 1D data extracted from the 2D original fingerprint image in a specific direction. However, there might be a possibility that the FRT's order in each line of the fingerprint image is revealed, because the extent of the intensity FRT is dependent on the FRT's order.In this paper, to solve this problem, we propose the fingerprint templates generated by the DFSTs, and indicate the accuracy of the fingerprint recognition by use of the proposed templates. Specifically, first, we compare the amplitude distribution (AD) and PD of the discrete FRTs (DFRTs), the discrete fractional cosine transforms (DFCTs) [14] and the DFSTs [14] of the fingerprint image, and decide the most appropriate distribution and transform used in the generation process of the proposed fingerprint templates. Next, we evaluate the recognition accuracy of the generated fingerprint templates from the viewpoints of the ROC curve, which denotes the relationship between the false acceptance rate (FAR) and the false rejection rate (FRR), and the MER, where the FAR and the FRR take the same value [15]. In the analysis, the effects of the misalignment in the recognition process are not considered. Finally, we quantitatively analyze the robustness of the templates.

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

“…In our method, the 1D DFRTs with different transforms' orders were applied to the ID data extracted from the 2D original fmgerprint image in a specific direction. Moreover, I have also proposed the fingerprint templates based on the DFCT and the DFST [14] and evaluated the properties [15]. In the evaluation, however, the genuine distributions were obtained from the fingerprint templates generated from the fingerprint data to which Gaussian random noise was added.…”

EER, Robustness and Generation Condition of Fingerprint Templates Based on the Fractional Cosine and Sine Transforms

“…In our method, the 1D DFRTs with different transforms' orders were applied to the 1D data extracted from the 2D original fingerprint image in a specific direction. Moreover, the fingerprint templates based on the DFCT and the DFST have also been proposed and evaluated the properties [12], [13]. In these studies [11]- [13], however, the genuine distributions were obtained from the fingerprint templates generated from the fingerprint data to which Gaussian random noise was added, though the genuine fingerprint images should have been used in nature.…”