SVM-based Fingerprint Classification Using Orientation Field

Abstract

This paper presents a classification method of fingerprint using orientation field and support vector machines. It estimates orientation field through pixel gradient, then calculates the percentages of the directional block classes. These percentages are combined as a four dimensional vector, by which the trained hierarchical classifier classifies the fingerprint into one of the six classes it belongs to. Experiments show that this method has high classification accuracy as well as low computational time cost.