Towards Adaptive Biometric System: Updating the Enrolled Template

Biometric authentication offers enhanced security and user convenience. However, variations in biometric sample quality and changes in biometric traits over time present significant challenges to maintaining accurate recognition. Adaptive biometric systems, which update enrolled templates to accommodate such variations, have emerged as a promising solution. This paper proposes two template update methods: one for iris recognition using a manually constructed decision rule and another for face recognition employing a lightweight neural network. The enrolled-template, which includes iris feature-vectors matching to iris photos taken under various lighting situations, is obtained using the iris technique. The face technique chooses the best feature vectors based on quality estimate, which improves recognition accuracy. Both methods demonstrate decreasing recognition error rates, emphasizing the potential of the adaptive biometric systems.