Theoretical statistical correlation for biometric identification performance

Measurement and evaluation of biometric device performance is critical to end users and consumers of these devices. In this paper we present explicit theoretical correlation models when the biometric matching process is stationary that can be used to derive variance estimates of biometric performance metrics. We focus in this paper on the failure to enroll, the failure to acquire, the false match rate and the false non-match rate. We further present unified notation that makes our correlation frameworks possible.