Who’s Who?

Hank Hogan

Many secure biometric methods of identification appear ready to do the job. However, technical and practical hurdles must be overcome before the technology sees widespread use.

Biometrics isn’t just a good idea. It’s the law, as spelled out in the Enhanced Border Security and Visa Entry Reform Act of 2002 and various other regulations. Passed in May, the law mandates identification based on the measurement of physical characteristics such as fingerprints, retinal patterns and facial features. Biometric techniques promise identification that can’t be forged, forgotten or lost.

Many of these solutions depend on photonics technologies. In some cases, a special light source is needed. The main photonics role, however, lies in the area of sensors. Current sensor technology works, but there is room for improvement. An example is fingerprint scans, a widely used biometric technique.

“We’ve challenged the photonics industry to actually advance the state of the art in the CMOS space,” said Joseph Atick, president and CEO of Minnetonka, Minn.-based Identix Inc., which provides hand, fingerprint and facial biometric products. “Right now, for the high-resolution 500-dpi technology, we can’t use CMOS because the signal-to-noise is not good enough to get to the quality that is required by the FBI.”

A look at various other photonic techniques indicates the obstacles that must be overcome before taking the biological measure of a man or woman.

The rewards of meeting such challenges could be substantial. A September forecast by technology consultants Frost & Sullivan of San Jose, Calif., put the worldwide biometric market at less than $100 million. By 2006, it is projected to grow to $2.05 billion. That’s a compound annual growth rate of more than 85 percent for biometric hardware alone. The projection “does not include third-party revenues, so it doesn’t include systems integrator revenues or value-added resellers,” explained Prianka Chopra, an analyst with the consulting firm.

Seven biometrics technologies are included in that $2 billion figure: hand geometry; non-FBI-related finger scan; iris, retina and face recognition; and signature and voice verification. Photonics technologies play a substantial role in nearly all of these.

The estimate also reflects a significant shift in the focus of the industry. In years past, biometric solutions were marketed to large businesses and commercial enterprises. The proposed uses included control of access to buildings and identity verification for computer login. There was also a substantial market involving background checks for suspected criminals.

From private to public

Today, in contrast, the focus is on large public-sector applications, such as the biometric checks at border crossings, as envisioned by the 2002 act. That refocusing has implications for both the technology and the market. Michael Thieme, director of special projects for consultants International Biometric Group LLC in New York, noted that this concentration on the large public sector means that the market will be slow to develop. “These are long sales cycles in these types of government projects, but it’s definitely going to be the impetus for the growth that we project for the next few years,” he said.

As for the technology impact, that gets a bit complicated. The 2002 act and other legislation don’t specify the biometric approach or standard to be used, but they lay out a timeline. By late 2004, the Immigration and Naturalization Service (INS), for example, is supposed to install biometric devices at every port of entry into the US, with the ultimate goal to collect information from every non-citizen entering the country. That’s a tall order, given INS estimates that about 35 million people entered in 2000. According to Thieme, the various government directives call for implementing the biggest and most complex biometric effort in history.

Part of the complexity arises from fundamentals. All of these techniques have an enrollment phase, when a person is initially scanned and pertinent body characteristics are captured. This is often done with the greatest possible thoroughness and may involve an interview and quite a bit of time. It also may include comparing an individual against an extensive database of “undesirable” people, such as those on a watch list or with a criminal record.

Once the information is captured, it can be used for day-to-day verification by matching data with a real-time scan of the person. This establishes that the individual is who he says he is. Such a one-to-one match is easier and faster than the more thorough background search that may be part of the enrollment phase.

Put your thumb here

The primary factor used for identification, of course, has always been facial feature recognition, joined more than a century ago by fingerprint matching. They are still the most accessible means, and they represent years of accumulated data. Fingerprints are a staple of law enforcement, and this sector is a significant part of the overall biometrics market. Frost & Sullivan projects that fingerprint scanning, the leading biometrics technology, will continue to dominate the market.

However, it may not be photonics-based fingerprint identification. Silicon sensors are produced by such manufacturers as Infineon Technologies AG in Germany, ST Microelectronics in Switzerland, Sony in Japan and Veridicom in the US. These devices can capture fingerprint information via capacitance, heat or other nonphotonic means. They are compact and low in power, which is advantageous for applications using laptops and other portable equipment. These sensors are already showing up in high-end laptops and, according to Chopra, it won’t be long before every laptop will have a silicon-based finger-scan sensor. Thus, devices based on photonics technology will soon have a formidable challenger.

But silicon sensors may not be suitable for some applications. A very small sensor cannot easily pick up the minutia that distinguishes one fingerprint from another. “What will likely happen is that there is going to be room in the market for both silicon and optical approaches. I would be very surprised if silicon technology can be designed to be as robust as it needs to be for things like civil ID applications,” Thieme said.

As for innovations in optical methods, Atick points to two that were developed by his company. The first is a patented approach that provides improved moisture tolerance (Figure 1). One problem confronting a body scan is that skin may be damp with perspiration or water, so Identix has engineered an optical solution that penetrates moisture. The technology extracts information from wet fingerprints and turns smudged images into clear ones.

Figure 1. Identix Inc. produces a fingerprint-identification product that employs its patented ClearScan moisture-discriminating optics that eliminate loss of ridge and pore structure detail (left) and allow the use of moisture to improve coupling for worn and fine ridges (right).

The second innovation is a full-hand scanner. Unlike a finger, the palm of a hand cannot be pressed against a surface and flattened to make an image that includes its distinguishing features. In the past, the only solution was to put ink on the hand and press it against special paper coating a drum. Identix has created a device that optically images a hand from the tip of the finger to just below the palm. The 1000-dpi image is captured cylindrically so that the full features of the hand remain intact.

For fingerprint identification, the trend is toward higher resolution and lower cost. That is one reason why Atick would like to see better CMOS sensors. The company currently uses a mix of CMOS and CCD imagers, but that may have to change as market forces push imaging requirements higher. “Right now, we capture at 500-dots-per-inch resolution for the fingerprint … and we are OK with the price points.” However, he expects the market in the next few years to go to 1000 dots per inch.

Figure 2. Biometric identification is increasingly used for secure access to computer systems and communication equipment. Courtesy of Iridian Technologies.

A face in the crowd

The use of facial features for identification lags behind fingerprints but should grow substantially, Chopra predicted.

Unlike fingerprinting, there are no competing approaches to photonics-based imaging, but there also is no need for specialized sensors. Identix manufactures face-recognition systems, and Atick said that its customers use off-the-shelf cameras for image capture. Many of the systems developed by Imagis Technologies Inc. of Vancouver, British Columbia, Canada, have found a home in law enforcement agencies throughout North America and the world (Figure 3). Other companies are also actively pursuing the face-recognition market.

Figure 3. For facial identification, Imagis Technologies Inc.’s ID-2000 extracts information from reflected light in a 2-D image to develop 3-D information about a person’s face. A video camera captures the image, the system characterizes it with a mathematical process and the image is compared with others in a database until a match is found.

Dave Lutes, chief engineer with Imagis, said the company’s technology recognizes faces using a deformable surface model. The image is treated as a three-dimensional object, even if it is derived from a two-dimensional photograph. The computer model uses algorithms to accommodate differences in lighting position and intensity. In effect, the system mathematically crawls along a multidimensional surface, tweaking the lighting to get the best possible fit of model to image.

Enrollment for face recognition can be done by a camera specifically for that purpose or from existing photographs. The latter appeals to law enforcement because a picture database of people who have criminal records already exists. Face recognition also should provide positive identification for security purposes.

There are important constraints on the technique. “It doesn’t require … special lighting, but … the accuracy of the system will increase as your recognition environment approaches the same lighting conditions as your enrollment environment,” Lutes said.

The eyes have it

Another biometric approach that is expected to work well in the security market, such as the travel industry, is iris recognition. Iridian Technologies Inc. of Moorestown, N.J., is active in this area and touts it as offering a one-of-a-kind biometric identification solution. The company literature says that the human iris is unique, even to individual twins and to a person’s right and left eyes (Figure 4).

Figure 4. The uniqueness of the iris makes it a prime candidate for personal identification. Iridian Technologies’ IrisAccess 2200 detects an individual when he or she is within 3 to 10 in. of the imager. The camera captures the iris image, digitally processes it into a 512-byte IrisCode record and grants access to an area, a computer system or a videoconference only when a match is made.

Although measuring only 11 mm or so across, the iris offers the advantage of being accessible and yet protected from the outside environment. Furthermore, the image of the iris is relatively insensitive to the angle of illumination, said Cambridge University researcher John Daugman, a pioneer in the field.

Company officials say the recognition process takes only a few seconds. As reported by the National Geographic Society in April, the technique was used to confirm an Afghan refugee girl’s identity.

moving target behind a curved and wet reflecting surface. To overcome these difficulties, iris recognition makes use of the near-infrared, in the 700- to 900-nm range. Iridian’s products use LEDs for illumination and a CCD sensor for image capture.

“If you just look at your eye in the mirror, for instance, you’ll see a lot of reflections of lights and windows, which cover up the iris. We can remove those reflections, essentially, by cutting out the daylight and providing our own controlled illumination,” commented Ulf Cahn von Seelen, director of algorithms for Iridian.

Retina recognition

The last biometric method involves the retina. As with the iris, the pattern of blood vessels in the retina is unique to each individual. In the past, capturing that pattern has required bright lights, pupil dilation, a skilled operator and cumbersome procedures. Retinal Technologies Inc. of Boston is one of the companies that has made use of modern optics and electronics to eliminate the need for such invasive methods. Greg Heacock, vice president of research and development, noted that it is important to choose the right target: the optic head found within the retina itself. This circumvents the problem of pigmentation differences among people.

“In all cases, the optic nerve head is white, and the vessels on top of it are red. It’s fantastic,” Heacock said. “You have 256 gray scales in one mode and essentially zero in the other. So it’s a great high-contrast image.”

Because blood vessels are being imaged, the pulsing of the blood ensures that what is being scanned is a living eye rather than a superimposed template.

Safeguards against hoaxes are built into the other methods as well. Atick said that his company’s products include testing to make sure that living subjects — and not hijacked biometric templates — are being presented for verification. Such methods are not foolproof, and biometrics companies continue to work to improve them. Efforts also continue to increase accuracy and applicability. There is a need, for example, for a means to identify people who lack the body part that is to be scanned or who otherwise can’t be measured by a given technique.

What lies ahead

The future of biometrics remains unclear. On the one hand, demand and requirements for security and authentication are increasing. A great deal of money is being spent, both on biometric systems and the supporting infrastructure. Published estimates have put the overall cost of the 2002 act in the hundreds of millions of dollars.

Biometric solutions are more expensive than such familiar approaches as ID cards. There are also questions about standards, both nationally and internationally. Cooperation among nations is essential to smooth border entry and exit operations, but that will likely require an agreement on which biometric methods to use. Hammering out such accords may take time. Meanwhile, the cost and the lack of an agreement will hinder deployment efforts. The consensus among industry analysts is that the interest in biometrics will not translate into dollars until 2003 and beyond.

However, many basic technological problems have been solved. As Thieme noted, “With the exception of extreme circumstances where you are pushing the limit of what biometrics can do — matching truly massive databases or working with noncooperative subjects — the technology is accurate enough to get the job done.”