Establishment of baseline performance for “end to end” biometric system evaluations

Usually biometric evaluations have been carried out with the aim of obtaining the performance of biometric algorithms using databases. Recently these evaluations have been broadened to test the performance of the overall system, conducting both scenario and operational evaluations which assess the end-to-end performance of the biometric system. Likewise, these evaluations are also being conducted for assessing the influence of other factors such as ambient conditions or usability on the biometric performance. In order to analyse if timing and/or error rates have been influenced by these parameters, it is essential to establish a baseline that allows to quantify the effect of such parameters. This entails determining the biometric performance considering two configurations: a standard configuration in which the biometric system will be evaluated under conventional conditions, and the evaluation configuration, in which the biometric system will be analysed with devices and test instruments that will be used for generating and measuring the evaluation conditions. Once the performance in both reference configurations has been calculated, it is possible to turn on all devices and instruments and execute the target evaluation, comparing the biometric system performance with all different environmental or usability conditions. In this paper, the definition of the above mentioned baseline will be given, and a a methodology is proposed for applying it to scenario and operational evaluations. This methodology is validated for a scenario evaluation by carrying out a fingerprint verification system considering the effects of temperature and humidity, and therefore, the influence of the climatic chamber used to generate and control these parameters.