A quality score honoring approach to semantic conformance assessment of minutiae-based feature extractors

Abstract

The use of biometric systems is steadily increasing, which leads to heterogeneity and, thus, possibly interoperability issues. In order to address such issues, standards are developed by ISO/IEC Joint Technical Committee 1. In the case of minutiae-based fingerprint recognition, for instance, ISO/IEC 19794-2 defines biometric data interchange records that vendors should adhere to. To validate adherence to this standard, ISO/IEC 29109-2 specifies means to perform syntactic conformance testing. Yet, a specification of semantic conformance testing is missing and is currently being discussed as a working draft amendment of ISO/IEC 29109-2. In order to contribute to this development, this paper proposes a general semantic conformance testing framework. Especially, a formal semantic conformance computation model is proposed that can be used to assess a biometric systems vendor's semantic conformance according to a ground-truth data set and to compare different testing approaches. Furthermore, an instance thereof is proposed that honors minutia quality scores when computing semantic conformance rates. Initial evaluation shows, that semantic conformance rates computed using this quality-honoring approach correlate with inter-vendor performance measures we would expect in a real-world scenario.