Fingerprint Identification with Combined Texture Features

The recognition using biometric fingerprint is low-cost and superior automated technique of verifying best match among two human fingerprints. In this paper, texture features are used for fingerprint matching. The matching methodology uses local scale and rotation invariant information. This paper proposes a novel method for fingerprint verification and identification by using combination of features of LBP, HOG and SIFT. The fingerprint image is first enhanced, segmented and rotated/scaled for different angles/values. The SVM and multi-SVM are used for verification and identification, respectively. The methodology is applied on database created by fingerprint sensor H520J00122 and also tested on FVC2004, FVC2002 database. The analysis points such as accuracy, time, TAR, FAR and FRR are measured. The proposed method also ensures the highest accuracy with moderate time of execution.