Semantic Conformance Testing Methodology for Finger Minutiae Data

This paper proposes a methodology to measure the semantic conformance rate of standardized biometric minutia interchange records. The paper proposes a fingerprint modality specific assertion test. A conformance test based on this methodology can attest for a given algorithm or software under test that the generated minutiae templates are a faithful representation of the input signal (i.e. fingerprint image). The test methodology is based on ground truth data that has been composed by dactyloscopic experts. As individual experts assessment yields slightly diverging coordinates a clustering algorithm is proposed that merges a set of manually placed minutia into one ground truth data set. The methodology is evaluated on ten-print fingerprint images and the NIST baseline minutia extraction algorithm.