Nonparametric Analysis of Fingerprint Data

This paper demonstrates that, for large-scale tests, the match and non-match similarity scores have no specific underlying distribution function. The forms of these distribution functions require a nonparametric approach for the analysis of the fingerprint similarity scores. In this paper, we present an analysis of the discrete distribution functions of the match and non-match similarity scores of the fingerprint data. This analysis demonstrates that a precise Receiver Operating Characteristic (ROC) curve based on the True Accept Rate (TAR) of the match similarity scores and the False Accept Rate (FAR) of the non-match similarity scores can be constructed without any assumption regarding operating thresholds and the forms of the distribution functions. The area under such an ROC curve computed using the trapezoidal rule is equivalent to the Mann-Whitney statistic directly formed from the match and non-match similarity scores. Thereafter, the Z statistic defined using the areas under ROC curves along with their variances is applied to test the significance of the difference between two ROC curves. Four examples from NIST’s extensive testing of commercial fingerprint systems are provided. The nonparametric approach presented in this article can also be employed in the analysis of other biometric data.