|
|
TECHNOLOGY
1. Fingerprint
Identification.
From the very soon AST bets for developing its own
technology. In the field of automated fingerprint recognition AST decides to
carry on a constant effort on R&D in software for processing and comparison
of fingerprint samples, reaching minimal processing time and high accurate at
identification. As a common characteristic, all biometric technologies are based
on biologic and morphologic features which define individuals in a unique way,
and make possible automated identification.
Besides all are based, at
identification process, on the comparison between the result of the processing
of a captured sample, commonly named biometric pattern or template, and another
one obtained as the same result of a sample prior captured and thus stored,
treated as a reference template. Global measure of that comparison is expressed
as a similitude in per cent: two biometric samples will be identical if that
similitude reaches 100%, and will be totally non-identical if similitude is 0%.
Both values are, obviously, extreme and limit, and thus unreachable. Nowadays
there are several biometric technologies to establish automated identification,
some of them available, some in an early stage.
Among them can be considered:
Iris recognition: although is reliable, it has the disadvantage that is
expensive and individuals seem to feel some invasive sensation when capturing a
photograph of the eye. Its time-response is not the more adequate generally, and
devices must be calibrated periodically. Besides, the biometric template is
onerous in size. Facial recognition: although this technology is not invasive
and has improved its performance in the last years, it is still two magnitude
orders in terms of reliability below fingerprint and iris. In fact the US
administration National Institute for Standards of Technology, NIST, recommends
the inclusion of facial in travel documentation and id cards, as a previous
filter, leaving the real identification to other technology, mainly fingerprint.
Hand recognition: Its great advantage is the minimal size of the obtained
template. However its accuracy is so low, and usually is used along with
fingerprint recognition in large criminal AFIS systems as a support.
To the
point of view of practical application no other technology is observed, except
of course automated fingerprint identification, due to several reasons as lack
of maturity, complex enrolment, cost, and low practical performance. For
instance, the recognition through analysis of genetic markers, commonly known as
DNA recognition, it is not included as an automated one because time response is
long and its costs prohibited.
Fingerprint recognition has been used from the
second half of the 19th Century until our days, and is largely documented and
huge volume of literature has been produced. As a result, a complete body of
knowledge is available, and from it the automated process has been developed.
For all of these reasons automated fingerprint identification is the most
available, costless, efficient, and reliable automated biometric technology
nowadays. In general terms individual fingerprint pattern is based on the skin
relief of the fingers of the hands which are formed along the first months of
gestation. It is the set of the ridges and valleys which basically end or
bifurcate. That relief forms well-defined structures that can be recognized by
an expert, human or machine. Now the steps of the automated fingerprint process
are resumed basically. First, an image of the fingerprint is obtained, from
certain capture device,
Next the system
decides where the quality image is concentred, in order the image to be
processed,
Next step is
to apply a threshold,
The system analyses it to get it
thinner,
And
finally the biometric pattern, the fingerprint template, is
extracted,
2.- Time Response in Large
Fingerprint Databases.
One of the main features where automated fingerprint identification is
based on is the easy is to search and to match fingerprints or latentprints. For
that is critical not only matching time, but also the global response time of
the system from it is queried until some result is supplied. In other words, a
high throughput must be granted, and it depends on the matching but also on the
way we look for.
Let’s suppose we want to search a fingerprint in a database including N
of them. If we assume that the matching time among the one we are looking for
and any in the database takes t seconds, we could expect that in the worst case
the time to find it would be,
That is, proportional to the size of the database. Most of Large Systems
follow that behaviour, and must deal with it with proprietary hardware, RISC
expensive solutions and parallel processing with onerous systems.
However, one of the working lines of AST has been to make independent
this time response from the size of the database. Next figure represents the
behaviour of AST fingerprint searching engine versus competitors:
From the chart it is clear that,
from a certain database size, AST time response is constant, while others
systmes still increase it.
3.- Information Security in
Automated Fingerprint Identification Systems.
Due to the sensitive and high security environment where AST carries on
its activity, adaptation, using, and design of secure information architectures
of systems have been essential.
AST provides secure technology for deploying large fingerprint projects
on potentially non secure networks, in order for them to become almost
impregnable. Next figure relates this background:
4.- Fingerprint Identification in
Smartcard , Match on Card, MoC, Algorithm.
Match on Card technology, MoC, lies on that the fingerprint verification
takes places inside an smartcard, using its processing means.
Thus a central
database can be eliminated since reference biometric templates are yet stored in
the holder’s id card.
By the
other hand reliability of matching is assured by the document’s vendor, since
matching algorithm is shipped with it. It is also guaranteed that biometric
information is always inside the card, and never is stored externally, what is
essential in some environments. Performance of AST MoC technology is: Mean time
of matching operations: 400 milliseconds. Mean time of non-matching operations:
800 milliseconds. Maximum time of matching operations: less than 1 second.
Maximum time of non-matching operations: less than 2 seconds. Regarding to the
card resources needed, AST has got them to be very cheap: 9 KB. 1 KB. RAM 2 KB.
With respect to reliability, this curve is obtained:
|
