Fingerprint Minutiae Matching Through Sparse Cross-correlation

In this paper, we introduce a novel minutiae-based matching algorithm for fingerprint recognition. The method is built on an elegant and straightforward mathematical formulation: the minutiae set is represented by a train of complex pulses and the matching algorithm is based on a simple crosscorrelation. We propose two different implementations. The first one exploits the intrinsic sparsity of the signal representing the minutiae set in order to construct an efficient implementation. The other relies on the Fourier transform to build a fixed-length representation, being thus suitable to be used in many biometric crypto-systems. The proposed method exhibits performance comparable with NIST's Bozorth3, that is a standard de facto for minutiae matching, but it shows to be more robust with cropped fingerprints.