Technologies within a Technology...
Since we have already discovered that, maybe not so surprisingly, biometric technology has been in
development and available since the eighties, it is probably also not so surprising that the technology
has not only extensively progressed since then, but that it has also become a much more complex
industry with many sub-divisions. In general terms, we understand that biometric technologies
can provide us with a means of creating a unique identification for an individual, through the
measurement of certain biological characteristics of that individual, perhaps utilising fingerprint ID,
or facial recognition software and equipment for example. However, when we look at the individual
biometric technologies themselves, it is particularly interesting to note that although they share
the same task and the same goal, the developers or manufacturers of, for example, fingerprint
ID technology products and equipment, do not share the same philosophy and methodology in
achieving that goal, and of course each believes their solution to be the most effective. In this issue we
begin to take a cross-section of the industry sub-divisions, starting with fingerprinting, to give an
insight into some of the different processing techniques applied, and some of the strengths and
weaknesses within each technology.
Digitised Digits
It is logical that we should start with fingerprinting, as it is reported to be both commercially
and technically, the most mature sector of the Biometrics industry. As a result there is more detailed
documented information concerning fingerprinting available, since there is an ever increasing number
of players involved. As competition increases these players are becoming less secretive about the
technical methodology or the interpretation of how their software or devices work, since in a competitive
marketplace they must use that information to prove to the world, and their potential future
clientele that their product is the best. Fingerprinting is certainly the biometric that is currently
most widely in use around the globe, accounting for over 50% of all biometric systems to date.
It has been particularly widely used in law enforcement, since the impetus on the development
of the technology began to be encouraged over a hundred years go, when in 1893 the British Government
accepted as fact for the first time that no two individuals have the same fingerprint. Since that time companies have strived
to develop the technology, perhaps leading us to the assumption that this sector of the technology is
nearing perfection. Unfortunately this assumption is somewhat premature, as it seems that each
new technological step forward brings with it a whole new set of dilemmas for the technicians to resolve.
Capturing the print
Initially fingerprints were captured by a simple inking technique producing a 2Dimensional image,
progression took us to the next level involving latent finger images, and then on to inkless 3D
scanning processes. Of these new processes, optical sensor scanning has been in use for the longest
period, but in addition we now also have capture via other types of sensor such as, pressure sensors,
thermal sensors, solid-state sensors and ultrasonic sensors. The quality and accuracy of the image
produced is dependent upon a number of factors, both technical, in terms of the parameters of
the equipment and the software, and biological in terms of the condition and orientation of the
finger being scanned. According to the experts, ultrasonic methods tend to produce the best quality
scans as they are able to demonstrate a high contrast between the contours of the fingerprint. Using
this method, as the finger is pulled over the scanner it is subject to a series of low level electronic
pulses which, depending upon the manufacturer, could be measuring any of the following characteristics
of the finger print distance between lines, depth of lines, points of line ends or junctions, number
of ridges, valleys, loops, arches or islands, and possibly also, in higher resolution systems, the
location of minute sweat pores. At the same time an additional sensor will measure the speed at
which the finger is being moved across the scanner. This enables the software to piece the information
back together to produce, hopefully, an accurate image.
Comparison Techniques
Once captured, either manually or through specifically designed software programmes, the images
can then be used to compare against an existing print contained on a smart card (1:1) or against a
whole database (1:N) for example. The larger the database the longer the comparison time, but many
leading suppliers of fingerprint ID equipment claim checking times of less than 1 second, even when
matching 1:N on huge data bases containing hundreds of thousands of images. Since the early
days of taking ink fingerprints various comparison methods have been developed to discriminate
between prints, and today most new comparison methods are still based on software algorithmic versions
of those manual comparison methods originally devised many years ago.
Fingerprint Classification
Several companies are working along the premise that fingerprints can be categorised, as there
is some consensus of opinion that indicates all fingerprints fall into one of five or seven
different classifications (depending upon which philosophy you care to follow) of fingerprint-type.
Hence, in theory, when checking a particular print against a large database, the speed of the
verification process should be dramatically improved, as approximately 4/5 (or 4/7) of the database could
be immediately eliminated as non-matching by simply categorising the print in question. This seems to
provide a spectacular breakthrough, but unfortunately it still needs some perfecting.
The problem is that there are no really accurate or specific guidelines laid down relating to
the parameters of each category, and in tests of large numbers of prints, experts could not agree on
the category of around 17% of the prints involved. Th is, surely, would be an unacceptable level of
ID classification for any system. However, from past experience of watching in the wings as the
experts develop their new technologies, it must be true to say, that if the scientists have come
thus far with this theory, then given time and investment, we will almost certainly see this
methodology developed to new heights in the not too distant future.
 |
|
The interest in biometric solutions is now very significant and the intensive product
development within the fingerprint identification sector confirms the increasing
market demands.
|
Minutiae or Pattern?
The comparison techniques most widely used today are the various minutiae evaluation techniques
and the pattern based techniques. The first, minutiae based technique, allocates a set of
co-ordinates to a predetermined number of mapped geographical points on the fingerprint, such as
line junctions, called bifurcations, the locations of sweat pores, or where ridges and valleys
begin or end. These co-ordinates are then transferred into a formula for calculation, which
produces a unique multi-digit ID number for each fingerprint that can be compared to the nearest
matching number for 1:N or a similar figure within a certain set of parameters when matching 1:1.
Parameters have to be set and agreed, as there is a good possibility that the calculation
will not arrive at exactly the same multi-digit ID number every time for a particular fingerprint,
due to a number of potential variables such as, changes in the orientation of the finger, changes
in the condition of the finger, electronic noise in the software,
and the speed of movement of the finger.
The second most popular method is image or pattern based, which involves taking the captured
image and overlaying it onto an existing image to determine the levels of similarity. This process
is obviously much slower when comparing 1:N for identification, but is adequate for comparisons
of 1:1 for verification against a smart card or chip device containing only one print. There
are also several other methods, which are based on variations of the methods described above and
involve algorithmic calculations for the number of ridges, distance between ridges or pattern types of
the fingerprint under scrutiny.
The Pros and the Cons
As with any technology, there are both advantages and disadvantage in favour of, or against,
fingerprinting techniques, but, as you might expect, considering this is currently the most
widely utilised biometric the strengths carry more weight than the weaknesses, as the main
strengths are related to practicality, adaptability, ease of use, and of course, the one thing
every customer asks about first these days cost!
Advantages
• Tried and tested technology long history of usage compared to other biometrics.
• Fingerprint details are permanent.
• Cost effective solution fingerprint reader costs have reduced drastically from around 100
USD a few years ago to around 10 USD today.
• Reasonably acceptable accuracy levels compared to other biometrics and in relation to cost.
• Non-intrusive, simple and quick to use.
• Numerous potential applications.
• Easy to miniaturise for embedding into other equipment - for example, mobile phones, ATMs,
PCs and laptops, smart cards.
Disadvantages
• Cannot be used for covert applications such as surveillance needs full cooperation from
the user to function properly.
• Stigma still exists associating fingerprinting with criminality
• Small number of the population cannot use due to disability.
• Poor recognition may occur due to temporary changes to a print eg cuts, bruises, disease,
dirt or worn out prints.
• Poor recognition may occur due to the size and orientation of the scanned fingerprint sample compared
to the already filed image.
There is huge international interest being generated in this technology, at almost every level and
across many industry sectors where fingerprinting could be integrated relatively easily and cost
effectively into existing security or access control measures. As a result there are a growing
number of companies, of all sizes and stature, from every corner of the globe, that are now
specialised in the design, development, manufacture and integration of biometric fingerprinting
technology. Some are relatively small players, but some are industry giants, and some have
only recently become involved in the field, whereas others have been in the business for over 30
years. Below we have named just a sample selection of current players for information:
• NEC Electronics Corp USA
• IDTECK Co. Ltd Korea
• ISL Informer Systems Ltd UK
• Lightuning Taiwan
• BMF Corporation Japan
• Identix UK
• Digital Persona USA
• Fingerprint Cards Sweden
• Idencom Germany
• Neurotechnologija Ltd Lithuania
• SAFLINK Corp USA
• Technoimagia Co Ltd Japan
• TSSI UK
In part 5, the final part of our Biometrics Academy we will cover the principles behind some of
the other biometric technologies such as facial recognition, keystroke dynamics, gait recognition,
iris and retina recognition, voice recognition and hand and mouse written signature ID.
|