Recognition Problem in Biometric Data Studies: Nonparametric Dependence Characteristics and Aggregated Algorithms

Published

September 14, 2006

Author(s)

Andrew L. Rukhin

Abstract

Biometric systems designed to detect or verify a persons identity are widely used in homeland security. A variety of commercially available biometric systems are now in existence; current technological progress makes it possible to evaluate these systems consistently and comprehensively. Recognition or identification problem of biometrics is important for such evaluations. In identification systems, a biometric signature of an unknown person, a probe, is presented to a system, which compares the new signature with a database of biometric signatures of known individuals. On the basis of this comparison, the system reports the similarity scores of the probe to the signatures in this database, called the gallery. The gallery items are then ranked accordingly to their similarity scores of the probe; the top matches with highest similarity scores are expected to contain the true identity. This work addresses two following issues: how to compare algorithms on the basis of their similarity scores for face recognition and how to combine different algorithms. An example from the FERET (Face Recognition Technology) program with four face recognition algorithms is examined.

Citation

Statistical Methods in Counter-Terrorism

Pub Type

Others

Download Paper

Local Download

Keywords

Aggregated algorithm, Copula, Correlation, Gallery, Metrics on permutations, Probe, Permutation matrix, Similarity score

Citation

Rukhin, A. (2006), Recognition Problem in Biometric Data Studies: Nonparametric Dependence Characteristics and Aggregated Algorithms, Statistical Methods in Counter-Terrorism, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=150387 (Accessed April 24, 2024)