Abstract
When creating a biometrics identification system, it is important to verify that the estimated performance for the system satisfies the target performance for the required matching precision and response speed. As a result, a model that can quantitatively assess the performance of the identification system is essential. The performance of an identification system using identification (oneβtoβmany matching) is different from the performance of an identification system using oneβtoβone matching. Performance varies not only with the rate of false nonmatches and the rate of false matches, but also with the distribution of the number of searches and the number of search interruptions. In this paper, the authors create a model to assess performance in a matching system overall. In addition, by applying to a matching performance model the search matching algorithm (highβspeed) proposed by the authors and a general linear search algorithm, the authors are able to clarify the effects on search matching precision of the distribution of the number of searches, the number of search interruptions, and the number of registration data records. Β© 2005 Wiley Periodicals, Inc. Syst Comp Jpn, 36(11): 60β69, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.20310