Fingerprints: Fixed Length Representation via Deep Networks and Domain Knowledge
This addresses the need for efficient and accurate fingerprint matching in biometric systems, offering a novel representation that improves performance over commercial methods.
The paper tackled the problem of representing fingerprints with fixed-length features instead of variable-length minutiae sets, achieving higher search accuracy (97.9% vs. 97.3% Rank-1 accuracy on NIST SD4) and significantly faster matching speeds (682,594 vs. 22 matches per second).
We learn a discriminative fixed length feature representation of fingerprints which stands in contrast to commonly used unordered, variable length sets of minutiae points. To arrive at this fixed length representation, we embed fingerprint domain knowledge into a multitask deep convolutional neural network architecture. Empirical results, on two public-domain fingerprint databases (NIST SD4 and FVC 2004 DB1) show that compared to minutiae representations, extracted by two state-of-the-art commercial matchers (Verifinger v6.3 and Innovatrics v2.0.3), our fixed-length representations provide (i) higher search accuracy: Rank-1 accuracy of 97.9% vs. 97.3% on NIST SD4 against a gallery size of 2000 and (ii) significantly faster, large scale search: 682,594 matches per second vs. 22 matches per second for commercial matchers on an i5 3.3 GHz processor with 8 GB of RAM.