Measuring system performance is conceptually straightforward; it is the interpretation of the results and their use as predictors of future performance that are the exceptional challenges in system evaluation and the experimentation in general. Good experimental design is critical in evaluation, but there have been very few techniques that a scientist may use to check their design for either overlooked associations or weak assumptions. For biometric and vision system evaluation, the complexity of the systems make a thorough exploration of the problem space impossible. This lack of verifiability in experimental design is a serious issue. In this paper, we present a new evaluation methodology that aids the researcher in discovering false assumptions about the homogeneity of cofactors - when the data is not well mixed. The new methodology is then applied in the context of a biometric system evaluation.
Citation: NIST Interagency/Internal Report (NISTIR) - 7156Report Number:
NIST Pub Series: NIST Interagency/Internal Report (NISTIR)
Pub Type: NIST Pubs
biometrics, evaluation, performance characterization, sample design, variance estimation