Investigating the reliability of positive user recognition based on multiple fingerprint images

Fingerprint is a biometric not only widely used in forensic investigation, but also in civil applications for user access control. At a crime scene an investigator may find fingerprints inadvertently left by a suspect. Even though fingerprints of this kind can be very useful for finding the perpetrator, recent years have seen multiple cases of misidentification caused by partial or latent fingerprints. Question about the usefulness of fingerprints in courtroom is raised due to these incidents. In civil applications a fingerprint identification system requires a user to present a live finger for image capturing at registration. Upon recognition new images from the same finger are captured, processed, and then matched against the registered template. During the processes, how often would this user be positively identified with the newly acquired images? In this paper we conduct simulations with a few fingerprint databases containing multiple fingerprint images from every finger and report our findings. Effects of minutia quality and threshold on matching are also discussed in this paper.