Fingerprint recognition based on spectral feature extraction

With the advancements in computational techniques and computer technology, biometric-based authentication has a strong potential to be widely utilized in a variety of applications. Among various biometrics, fingerprint-based identification is the most mature and widely accepted technique. In this paper, a fast procedure that exploits the spectral features of the fingerprint to obtain a compact descriptor representation, that is both rotation and translation-invariant, is presented. Experimental results have shown high matching accuracies that are equivalent or even better than those reported in the literature.