An automation for robust design of multimodal biometric systems

This paper presents a continuous investigation into performance evaluation of multimodal biometric systems. In the recently completed work, deviations of matching scores are introduced to stand for uncontrollable noises when analyzing different combinations of data fusion methods and normalization schemes. This paper further refines the systematic approach of performance evaluation with automated processing. It proposes a framework that is scalable when combining different biometric databases into a larger subject pool. The developed application allows users to identify a larger set of deviation values for noises, to automatically generate test cases for all biometric modalities, and to use a set of graphs and reports that are in tune with the common industry (commercial) standards as opposed to purely numerical outputs. In addition to technical details, this paper also includes results from experiments.