Abstract:
The aim of this paper is to show the performance of a complete scheme of the minutiae-based fingerprint automatic verification system, previously proposed [13], and recently improved, evaluating its performance in the verification mode with the large MCYT database [10]. The design of this database, the characteristics of the stored fingerprint images, the minutiae extraction procedure, and the pattern recognition process, are briefly described in the first part of the paper. Subsequently, several experiments are presented in order to analyse the influence on the system performance of a variability factor, like fingerprint position variability during acquisition stage. The use of multiple-reference strategy and score normalization technique have permitted us to highly improve verification performance.
