A fingerprint matcher can make two types of errors: a false match, in which the matcher declares a match between images from two different fingers, and a false nonmatch, in which it does not identify images from the same finger as a match. A system’s false match rate (FMR) and false nonmatch rate (FNMR) depend on the operating threshold; a large threshold score leads to a small FMR at the expense of a high FNMR. For a given fingerprint matching system, it is impossible to reduce both these errors simultaneously.It also lists some US government's systems' performance and the FNMR seems to be higher than FMR in just about all system. Does this means that the system will not match more often than finding a match?
Fingerprint identification system performance is measured in terms of its false positive identification rate (FPIR) and false negative identification rate (FNIR). A false positive ident i f icat ion occur s when the system finds a hit for a query fingerprint that is not enrolled in the system. A false negative identification occurs when it finds no hit or a wrong hit for a query fingerprint enrolled in the system. The relationship between these rates is defined by FPIR = 1 - (1 - FMR)^N, where N is the number of users enrolled in the system. Hence, as the number of enrolled users grows, the fingerprint matcher’s FMR needs to be extremely low for the identification system to be effective. For example, if an FPIR of 1 percent is required in a fingerprint identification system with 100 million enrolled users, the FMR of the corresponding fingerprint matcher must be on the order of 1 in 10 billion. Such a stringent FMR requirement can usually be met only when fingerprints from all 10 fingers of a person are used for identification. This explains the need to continuously decrease the error rates of fingerprint matchers employed in large-scale identification systems.
I always want to know why my Fujitsu P1510's fingerprint log in system always rejects me with a rate like 20-25 tries to get one successful log in (I have not disabled that system). This paper offers some plausible explanation:
Fingerprint sensors embedded in consumer electronic devices tend to have a smaller sensing area. This factor, combined with users’ improper placement of their finger on the sensor, results in a limited overlapping area between two impressions of the same finger, as Figure 5c shows. Given the very small number of minutiae in the overlapping area, it is difficult to determine if two fingerprints are from the same finger.
One way to alleviate this problem is to utilize level 3 features to improve the matching accuracy in cases where there is only a small overlapping area between the two impressions. However, level 3 features may not be suitable for commercial applications because the sensors used in such applications usually provide only low-resolution images.
Why do I have so much trouble with fingerprint recognition system? This paper offers these suggestions:
In some cases, a fingerprint recognition system may not even successfully capture the user’s fingerprint. Failure to enroll (FTE) and failure to acquire (FTA) refer to the fraction of users who cannot be enrolled or processed by a particular system due to the poor quality of their fingerprints— for example, people such as manual laborers or the elderly with “worn-out” fingers. In practice, FTE can be rather high (a few percentage points) depending on the target population and the occupation of users in the population.
Due to nonideal skin conditions, inherently low-quality fingers, and sensor noise, a significant percentage of fingerprint images are of poor quality. Extracting features from and matching low-quality fingerprints, like those shown in Figures 5a and 5b, is a challenging problem that will require significant research.
Pressing soft finger skin on a sensor always introduces some distortion, which is generally not repeatable. Matched fingerprints may appear very different under severe distortion, as Figure 5d shows.
The paper's conclusion is reassuring to know that it is not totally my fault:
Although fingerprint recognition is one of the earliest applications of pattern recognition, the accuracy of state-of-the-art fingerprint-matching systems is still not comparable to human fingerprint experts in many situations, particularly latent print matching. Significant advances require not only a deeper understanding of friction ridge formation, but also adaptation of new developments in sensor technology, image processing, pattern recognition, machine learning, cryptography, and statistical modeling. While successful commercial applications have driven fingerprint-matching technology, more breakthroughs could be achieved with greater investment in fundamental research.