快猫短视频

Whose finger on the button?

鈥淏ESIDES taking electronic pictures, what else is an imaging chip useful
for?鈥 asks Larry O鈥橤orman. The answer is surprising and revolutionary. Today,
imaging chips give computers a sense of vision. But with a few modifications
these chips can add another dimension to a computer鈥檚 sensory repertoire: touch.
With this extra sense computers will be able to recognise people from the mere
brush of a finger.

O鈥橤orman is developing chips that recognise fingerprints. And he is not the
only scientist thinking this way. He and several competing teams have already
built prototypes that recognise humans by touch, and the first devices should
appear on computer keyboards within a few months. In a world obsessed with
security, the applications are obvious. The sensors will become silicon security
managers, says O鈥橤orman. One day they will sit unobtrusively on every computer,
cashpoint machine and car door, quietly managing who has access to what.

Driven by the need to identify criminals, police forces have built up huge
databases of prints taken by inking a finger and making a print on paper. With
the advent of powerful computers and image-recognition software, these prints
and any new ones could be scanned into a computer. But this is a fiddly,
time-consuming and expensive process that requires bulky equipment. In the past
few years, optical scanners have become available that take a direct snapshot of
a fingerprint and desktop computers have become powerful enough to do the
recognition that was once performed by humans. So fingerprint sensing is nothing
new, but the devices require a light source, lenses and a way of converting the
image into digital format, and so cannot be made much smaller than a mobile
phone.

In 1995, scientists at Lucent Technologies, a research and development
telecommunications company based in New Jersey, came up with the idea of making
an array of pixels sensitive to touch. If enough could be packed onto a chip,
such an array could pick out the ridges and valleys of a fingerprint. Lucent鈥檚
touch sensitive technology relies on the ability of conductors to store charge,
a property known as capacitance. A capacitor consists of two conductors
separated by a type of insulating material known as a dielectric. Importantly,
the capacitance is inversely related to the distance between the conductors. So
conductors that are close together have a higher capacitance than ones that are
further apart.

Each pixel in the Lucent design is one tiny conducting plate covered with a
dielectric film. When in contact with a finger, the skin acts like another
conducting plate forming a crude capacitor. And since the ridges on a
fingerprint will be closer to the pixels than the valleys, measuring the
capacitance at each pixel in the array produces an electronic snapshot of the
fingerprint. In 1996 Lucent and a group of venture capitalists set up a company
called Veridicom based in Santa Clara, California, to bring the idea to market.
O鈥橤orman is its chief scientist.

With conventional chip fabrication techniques, Veridicom packs some 90 000
pixels into an array the size of a postage stamp. This produces enough
resolution to map the characteristic patterns of whorls and bifurcations that
image recognition software relies on for identification. The distance between
valleys and ridges differs by a factor of roughly two, depending on a person鈥檚
size and genetic make-up, says O鈥橤orman. To give an idea of scale, around four
pixels on the chip span the distance from ridge to ridge in the smallest
prints.

This unique approach has its challenges. For a start, humans can carry around
enough static electricity to fry the delicate components on a chip. So
Veridicom鈥檚 design includes ground wires that carry this charge away. An exposed
chip is also liable to become scratched, rendering the device useless, says Dave
Inglis, one of the chip designers at Lucent. At best, scratches change the
capacitance of the device by making the dielectric film thinner. At worst, they
break through the film, causing a short circuit. In addition, the skin鈥檚 oily
secretions can be corrosive and any ions they contain can migrate into silicon
structures changing their electrical properties.

The solution that Veridicom came up with is a thin coating of titanium
nitride and silicon nitride. Although only a few hundredths of a millimetre
thick, these layers block the migration of ions and protect the components from
corrosion. Silicon nitride provides the dielectric properties required for a
capacitor while titanium nitride, an extremely hard material, provides
robustness and resistance to abrasion. At one demonstration, people were invited
to attack the surface with a screwdriver鈥攖he surface was unharmed.

Veridicom is not the only company in the race to place fingerprint sensors on
every computer keyboard. Thomson-CSF, a French defence company, has developed a
chip that senses the heat difference between the ridges and valleys on a chip.
This is typically less than 0.1 掳C. The device consists of an array of
pixels that convert heat into an electric signal. 鈥淭he technology comes from
infrared imaging chips,鈥 says John Harris at Thomson-CSF.

The difficulty with this design is that the chip rapidly heats up when it is
in contact with the skin. As thermal equilibrium is reached, the ridges and
valleys in the image simply blur together. To get around this, users must brush
their finger across the chip while the processor takes several snapshots
(see Diagram)

.
The computer assembles these to form an image of the print.

Computer reconstructs snapshots to form image

A finger sensing chip
Sensor chip takes 50 snapshots

The sensor itself is only one half of the problem. Reliably identifying the
image that it produces is also crucial. 鈥淎 fingerprint image is one of the
noisiest image types,鈥 says O鈥橤orman. Fingertips become dirty, cut, scarred,
worn and creased, and may be dry or wet, so the images have to be cleaned up and
enhanced by maximising the contrast between ridges and valleys. The software
then looks for features called minutiae鈥攖he endings and bifurcations of
ridges and compares these to a set of templates of registered users in its
memory. Using these data, identification is usually possible within a
second.

For the moment, both types of chip require separate processors to carry out
the pattern recognition. This is why computers and mobile phones that have
built-in processors of their own will be the first to get fingerprint sensors,
probably later this year. Beyond that the goal is to build the processing power
directly into the sensing chip. By the standards of today鈥檚 chip makers,
Veridicom鈥檚 chip is sparsely populated鈥攊t contains 90 000 pixels compared
with the several million transistors on a Pentium chip. 鈥淲e have room to put a
lot of extra processing power on board and this is our goal,鈥 says O鈥橤orman. He
expects to have developed the all-in-one design within a couple of years.

So how well does Veridicom鈥檚 chip work? At a conference on solid-state
circuits in San Francisco earlier this year, Inglis and his colleagues reported
that the sensor accepts fewer than one in a hundred unauthorised fingerprints.
This is good enough for low-level security. 鈥淲e鈥檙e not protecting nuclear
weapons here, just your notebook computer,鈥 says Tom Rowley, Veridicom鈥檚 chief
executive.

Nevertheless, Harris believes the heat sensor offers protection against some
kinds of fraud that would fool existing optical scanners. 鈥淟atex prints that
slip over your own could fool an optical scanner, but not ours,鈥 he says. Then
there is the macabre prospect of criminals stealing fingers rather than cash or
credit cards. Since a dismembered finger would soon become cold, a temperature
sensor on the chip might help, says Harris. 鈥淏ut determined criminals could warm
the finger under their armpits,鈥 he adds. Basant Khaitan, an electrical engineer
at Veridicom, believes a dismembered finger would have different electrical
characteristics from an attached finger and so could be spotted. But Harris says
a line must be drawn: 鈥淵ou have to decide what level of security you want at
what price.鈥

At only $50 each and with prices set to drop, O鈥橤orman and Harris
expect fingerprint sensors to become ubiquitous within the next few years. This
could lead to a fundamental change in security and the way we approach it.
Today, carefully memorised passwords and pin numbers are an everyday method of
accessing bank and credit card information, protecting e-mail and electronic
letters from prying eyes and tracking users of websites and databases. But the
entire system is at the mercy of the painful frailty of human memory. Cheap
fingerprint recognition will make passwords redundant. In future, they say, the
key to your privacy will lie literally at your fingertips.

More from 快猫短视频

Explore the latest news, articles and features