Master Assignment
Estimating higher order FNMR/FMR curves
Type: Master EE
Period: T.B.D.
If you are interested please contact :
Description:
In many applications it is important that the administration of individuals does not contain any duplicate identities. Biometrics are very well suited for finding duplicate registrations in large datasets because biometric information is closely linked to an individual. When dealing with larger populations (e.g. all the citizens of a country) is important that the used biometric information has sufficiently high accuracy. Fingerprints are very well suited to search in large datasets because the accuracy of a single finger is high and on top of that it is possible to combine (fuse) the information of up to ten fingers of a single individual. A system using fingerprints to search for individuals is called an Automated Finger Identification System (AFIS).
As in any biometric system, in an AFIS there is a trade-off between the probability that a subject is found (related to the False Non-Match Rate (FNMR) of the AFIS) and the probability of generating false hits or fake duplicates (related to the False Match Rate (FMR)). The operating point in this trade-off is selected by setting matching thresholds based on estimated FNMR and FMR curves.
Goal of the research:
For a single finger, the FNMR and FMR curves are normally estimated using a representative dataset. For the curves of more than 4 fingers this becomes increasingly difficult due to the huge amount of data that is required to make accurate estimates of those higher order curves. The goal of this research is to develop a technique for estimating the higher order FNMR and FMR curves (5 to 10 fingers) based on measured curves of 1 through 4 fingers making it possible to predict the accuracy of the AFIS given a set of thresholds. The new technique should take into account dependency between fingers and produce estimates curves as well as confidence or credible intervals around the curves.
Profile of the student:
Some background in biometrics and interested in novel statistical techniques. Organized. Some experience in programming in for example Python or MatLab.