On Engineering Challenges of Applying Relevance Feedback to Fingerprint Identification Systems

Abstract

Effective fingerprint identification is critical in crime detection and many security related operations. This article investigates the use of Relevance Feedback with automatic fingerprint identification systems. It also summarises how several unique engineering challenges faced when applying relevance feedback are being addressed. Relevance feedback is a process for acquiring and using knowledge from a human user to improve the quality of results from an information retrieval system. It has been applied extensively to both text-based and images-based information retrieval systems, but not to fingerprint identification systems. Compared to automatic processes, relevance feedback has the potential to assist in faster convergence towards correct fingerprint identification and removes the reliance on black box fingerprint matching algorithm for the system's performance. Experimental results collected from a prototype software implementation confirm that relevance feedback can improve the quality of fingerprint identification queries significantly.