Different Feature Representation for Fake Fingerprint Classification

Abstract

Existing studies show the fake fingerprint detection faced a problem in dealing with a variety of materials that can be used to fabricate the fake fingerprints, type of sensor and the noises. These cause a lack of meaningful features extracted to represent the fake fingerprints. Continuous advancement in this domain does lead to the introduction of new materials for fake fingerprint fabrication. Meanwhile, the performance of classification shows a low accuracy when classifying the fake fingerprint fabricate by the unknown materials. Therefore, a good extraction method able to extract meaningful features is needed. This work aims to use two different based of features representation; pixel intensity-based and ridge length-based in order to gain the variant meaningful features. The idea is to represent both features gain from two different based on representation using the data statistical method, then select the best features before fused the features together.