Exploiting the “doddington zoo” effect in biometric fusion

摘要: Recent research in biometrics has suggested the existence of “Biometric Menagerie” which weak users contribute disproportionately to error rate (FAR and FRR) a biometric system. The aim this work is utilize observation design multibiometric system where information consolidated on user-specific basis. To facilitate this, database are characterized into multiple categories only belonging required provide additional information. contribution lies (a) selective fusion scheme invoked for subset users, (b) evaluating performance such two public datasets. Experiments multi-unit CASIA V3 iris WVU fingerprint indicate that fusion, as defined work, improves overall matching accuracy while potentially reducing computational time. This positive implications large-scale throughput can be substantially increased without compromising verification