A feedback paradigm for latent fingerprint matching

Latent fingerprints are of critical value in forensic science because they serve as an important source of evidence in a court of law. Automatic matching of latent fingerprints to rolled/plain (exemplar) fingerprints with high accuracy is quite vital for such applications. However, due to poor latent image quality in general, latent fingerprint matching accuracy is far from satisfactory. In this research, we propose a novel latent matching paradigm which takes feedback from an exemplar print during matching to refine the features extracted from the latent. The refined latent features are then used to update the baseline match scores and resort the candidate list retrieved from the database. Experimental results show that the feedback based matching mechanism improves the rank-1 identification accuracy of the baseline latent matcher by about 8% and 3% for NIST SD27 and WVU latent databases, respectively. The proposed feedback paradigm can be wrapped around any latent matcher to improve its performance.