Statistical significance as an aid to system performance evaluation

Abstract

Using forensic fingerprint identification as a testbed, a statistical framework for analyzing system performance is presented. Each set of fingerprint features is represented by a collection of binary codes. The matching process is equated to measuring the Hamming distances between feature sets. After performing matching experiments on a small data base, the number of independent degrees of freedom intrinsic to the fingerprint population is estimated. Using this information, a set of independent Bernoulli trials is used to predict the success of the system with respect to a particular dataset. © Springer-Verlag Berlin Heidelberg 2000.