Transitivity based enrollment strategy for signature verification systems

The enrollment phase of signature verification systems is a critical process in which reference data of a user is acquired that needs to be of satisfactory quality without overloading the subject by asking for too many repetitions. Many signature verification systems do not perform an enrollment quality evaluation at all, or only after capturing a fixed number of samples, accepting or rejecting the whole reference set. To limit the number of rejections and, as such, the false enrollment rate (FER), we propose a new algorithm for adaptive quality evaluation of samples during the enrollment process. This algorithm is based on transitivity criteria within a set of multidimensional reference vectors. We show that our approach leads to a significant reduction in the FER.