Fingerprint Classification Based on Orientation Field

Abstract

This paper introduces an effective method of fingerprint classification based on discriminative feature gathering from orientation field. A nonlinear support vector machines (SVMs) is adopted for the classification. The orientation field is estimated through a pixel-Wise gradient descent method and the percentage of directional block classes is estimated. These percentages are classified into four-dimensional vector considered as a good feature that can be combined with an accurate singular point to classify the fingerprint into one of five classes. This method shows high classification accuracy relative to other spatial domain classifiers.