Biometric features used in recognition systems are subject to aging. In addition, there may be variations in working conditions that are not taken into account when registering a user. Therefore, one of the tasks that must be solved when building a universal and long-term functioning biometric system is updating the biometric template. An algorithm for updating a biometric template is proposed that uses an estimate of the distance from the presented biometric features to the user's features registered under ideal conditions. Such an estimate is calculated by a neural network trained on a database with a large variability in registration conditions. The algorithm is tested on a face recognition system.
展开▼