Enhancement of robustness and accuracy of fingerprint recognition

Abstract

Identification systems using fingerprints are highly reliable and secure. A new and effective combined approach for fingerprint recognition is introduced in this paper. In the proposed method, fingerprint images are normalized in order to enhance the quality. Then, the orientation of ridges is estimated using gradient method. Afterwards, using Gabor filter, lost ridges are reconstructed. Due to the selective nature of the frequency and orientation of Gabor filter, and due to the fact that it has a favorable resolution in both spatial and frequency domain, it is used for reconstructing corrupted fingerprint images. Then, discrete wavelet transform and Gabor filter are jointly used for feature extraction. Being insensitive to rotation, Gabor filter is used for feature extraction. It is used in 6 different orientations in this study. Because discrete wavelet transform extract features in 5 levels, this method is also exploited for feature extraction. Finally, neural network is used for classification. The database used in the experiment is DB3 FVC2004 sub-database. The accuracy in the introduced combined approach is 99% while FRR, FAR, and EER are all calculated to be 1% which is better than those calculated for all investigated approaches in this paper.