Which computer configuration should I use?
What are the most common problems with fingerprints? Does everyone have a fingerprint?
Do I need the fingerprint images? What is a template? What is the size of a template?
What kind of fingerprint readers are available and which one is the best?
Can a fingerprint recover from cuts? Do they interfere with the identification?
What is the advantage of fingerprint recognition compare to other biometric methods?
Biometrics is the identification of a person based on his/her intrinsic characteristics, such as voice, writing, fingerprint, iris, face or hand geometry.
Identification is the technology that makes it possible to identify someone by their fingerprints. The person has to place his/her finger on the optical reader and the result will be the display of the person's name.
Verification is the technology that compares someone's fingerprint with the fingerprint of whoever this person claims to be. Because of that, this method requires also a password or an access card.
The Identification technology compares (in an efficient way) the requested fingerprint with all the recorded fingerprints in the system. The Verification technology compares the person's fingerprint with the already recorded fingerprint of this person. That is why the technologies are also called 1:N (one X all) and 1:1 (one X one).
For databases up to 1,000 fingerprints there are no big requirements. Any PC can be used, like a Celeron 1 GHz with 128MB RAM.
For databases up to 10,000 fingerprints it is recommended to use a dedicated PC to keep the 1 per second speed in the identification process. A proper configuration would be a Pentium 4 with 512MB RAM. A 800Mhz Bus is very important and makes a great difference in the performance.
For databases higher than 10,000 fingerprints it is important to use a multiple paralel servers architecture, the SpeedCluster. This way you will have many servers searching for the fingerprint, which makes the process faster.
In case of unique use of Verification method, a heavy server structure is not required. Only one calculation of a fingerprint is computed and then, it will be compared to a stored fingerprint. Unless many requests are made at the very same time, one server is enough; even a shared one.
Any biometric method may present some rejection problem, because they involve human and biological characteristics. That means that even a person whose fingerprint is already recorded may not be recognized. This is called "false rejection" and happens with any technology and manufacturer.
This problem rarely occurs (below 0.1% of the cases), but it is important to keep this possibility in mind during the implementation, so you can plan on what to do if that happens. The individuals that present this kind of situation are the elderly and children up to 6 years old. Some chemical products may also provoke the temporary reduction of a fingerprint quality. In addition, some people don't have fingerprints on some periods of the year, due to biological conditions associated to weather or to their own organism. In these cases alternative methods must be used, such as use of documents, passwords or access cards.
However, many false rejections happen because of an error during the registration, with the capture of a partial fingerprint, usually the tip or the side of it. That increases the possibility of a rejection, because the next time that fingerprint will be read, the captured image may be a different one, not registered yet. A correct registration is the best way of avoiding a false rejection.
The reader only captures a fingerprint image, which is processed by a software specially designed to extract some characteristic points, called minutiae. One regular fingerprint has about 50 minutiae. About 13 of them are needed to assure the identification.
When all the minutiae are extracted, they are put together. This is called a Template, the joining of all the extracted minutiae in a fingerprint. The identification is made by a triangulation process and geometrical relation between the minutiae only, not the entire image.
In the Fingerprint SDK, the templates usually have around 700 bytes.
The most used fingerprint readers fall into two types: optical and capacitive. The optical reader has a translucent prism or acrylic, while the capacitive one is opaque, a chip where you can put your finger on.
The optical models tend to present better image and bigger area to capture the image. The capacitive models tend to present worse image and smaller area to capture the image. However, some optical models are really bad, even worse than the capacitive ones.
Each application has a model thats is the most adequate. So you might want to consult us before choosing one or another. Like any other new technology, there are many myths about the readers. These myths can't resist some simple tests.
Except in some cases where the cut leaves a scar, a fingerprint recovers from a cut without problems, returning to its original aspect.
Even if there is a cut and a scar, the fingerprint will be easily identified, because 13 minutiae are needed for the process, while a regular fingerprint has about 50 of them.
Even identical twins have different fingerprints. Fingerprints have been used in identification for thousands of years and even today, with huge databases with millions of fingerprints, it hasn't been found one that is identical to another. Besides, each finger and toe has a completely different fingerprint from each other.
Fingerprint is the cheapest, fastest, most convenient and most reliable way to identify someone. That's why fingerprint alone has 2/3 of the biometric world market (according to an International Biometric Group independent report). And the tendency, due to scale, easiness and the existing foundation, is that the use of fingerprint will only increase. Cars, cell phones, PDAs, personal computers and dozens of products and devices are using fingerprints more and more.