Select your localized edition:

Close ×

More Ways to Connect

Discover one of our 28 local entrepreneurial communities »

Be the first to know as we launch in new countries and markets around the globe.

Interested in bringing MIT Technology Review to your local market?

MIT Technology ReviewMIT Technology Review - logo

 

Unsupported browser: Your browser does not meet modern web standards. See how it scores »

{ action.text }

High-resolution still images have been another factor in the improvement of face-recognition technology, in part because highly detailed skin-texture analysis has also become possible. With such analysis, any patch of skin–called a skin print–can be captured as an image, then broken up into smaller blocks that algorithms turn into mathematical, measurable spaces in which lines, pores, and the actual skin texture are recorded. “It can identify differences between identical twins, which isn’t yet possible using facial-recognition software alone,” Gross explains. “By combining facial recognition with surface-texture analysis, accurate identification can increase by 20 to 25 percent.”

What about the FRVT report’s claim that some face-recognition algorithms equal or exceed humans’ recognition capabilities? Phillips explains: “Humans are very good at recognizing faces of familiar people. However, they aren’t so good at recognizing unfamiliar people.” Since many proposed face-recognition systems would complement or replace humans, the FRVT’s comparative tests of the face-recognition capabilities of humans and software–the first such testing–were important for measuring the potential effectiveness of applications. Phillips says that at low false accept rates (a false accept rate is the measure of the likelihood that a biometric security system will incorrectly accept an access attempt by an unauthorized individual), six out of seven automatic face-recognition algorithms were comparable to or better than human recognition. These were algorithms from Neven Vision, Viisage, Cognitec, Identix, Samsung Advanced Institute for Technology, and Tsinghua University. Unfortunately, Phillips adds, “because the majority of FRVT 2006 participants haven’t disclosed the details of their methods, it’s not possible yet to assess what’s distinctive about these algorithms.”

How does the commercial payoff for face recognition look? Quite promising, because dozens of companies aim to cash in on face recognition’s potential as a biometric for credentialing and verification purposes. For the FRVT, venerable corporations like Toshiba and Samsung competed alongside companies like Neven Vision–just acquired by Google–and Viisage and Identix (which have just merged into L1 Identity Solutions), as well as alongside researchers from universities as diverse as Beijing, Cambridge, and Carnegie Mellon. What applications does a company like Google foresee for the technology developed by its recent acquisition, Neven Vision? According to a Google PR person, “We believe it offers promising integration possibilities with Google’s services, such as Picasa and Picasa Web Albums, particularly in terms of helping users organize and search their own photos.”

At Carnegie Mellon, Ralph Gross says that among other efforts, he and his colleagues have been “involved with local DMVs in order to scan images for driver’s licenses. I’ve gotten reports from the state level to say that, using face-recognition technology, they caught quite a number of people who applied for licenses in either different states or in the same state under a different name because their previous license got suspended.” It’s a growing trend. States using such technology include Massachusetts, Illinois, West Virginia, Wisconsin, Colorado, North and Southern Carolina, Oklahoma, North Dakota, Arkansas, and Mississippi. Nevertheless, Gross stresses, applying face-recognition technology to ID photos is a long way from having the capability that would let law enforcement search a city’s webcam networks for specific individuals. “With driver’s license photos, you have a controlled background, an operator telling you exactly how to position your face; the images are collected under comparable conditions. It’s much more restricted than the random-face-in-the-crowd problem, where you’re sticking a camera on a building.”

5 comments. Share your thoughts »

Credit: FRVT 2006 and ICE 2006 Large-Scale Results

Tagged: Computing, software, 3-D, facial recognition

Reprints and Permissions | Send feedback to the editor

From the Archives

Close

Introducing MIT Technology Review Insider.

Already a Magazine subscriber?

You're automatically an Insider. It's easy to activate or upgrade your account.

Activate Your Account

Become an Insider

It's the new way to subscribe. Get even more of the tech news, research, and discoveries you crave.

Sign Up

Learn More

Find out why MIT Technology Review Insider is for you and explore your options.

Show Me