On Controlling Genuine Reject Rate in Multi-stage Biometric Verification

摘要: An important problem in multi-stage biometric verification is to select an appropriate reject region. A region says which samples be rejected. Rejecting impostor does not incur any cost terms of user inconvenience, however, erroneously rejecting genuine leads both and administrator inconvenience. The becomes severe the applications that involve a huge number transactions. Such necessitate rate controlled. However, date, no work has studied on controlling (GRR) verification. In this paper, we focused GRR end, developed rejection method called symmetric method. Our adds following benefits verification: (1) it enables system control GRR, (2) allows calculate without estimation score distributions, (3) use assumption functional form distributions. We performed experiments two fingerprint datasets 6000 users face 3000 users. For data, achieved 18.96 percent 70.89 reduction EER by 1.5 9.4 scores for 3.27 85.83 0.3 14.4 scores.