Researchers see strides in biometrics

Georgia Tech scientists are studying ways to gather information about people from the way they walk.

Whether you stroll, stride, lurch or lumber, researchers at the Georgia Institute of Technology are studying ways to identify and track you by the way you walk.

The research, funded by the Defense Advanced Research Projects Agency (DARPA), aims to use radar and computer vision to create a unique signature based on a person's gait, along with leg and arm movement.

The resulting technology could be used to aid police in locating suspects by scanning large crowds of people for those who have a particular manner of walking, said Gene Greneker, principal research scientist at the Georgia Tech Research Institute.

"The way we walk can be pinned down to a particular person in a high percentage of cases," he said. "If we have this person in the file and we see them again, we can say that's them."

Spurred by the terrorist attacks of Sept. 11, 2001, law enforcement agencies and private companies have been looking for new technology tools. They've given high priority in particular to so-called biometric systems that recognize people by unique physical features, such as fingerprints, irises and the shape of the face.

The Georgia Tech researchers are looking to find a unique, but reproducible, signature in a person's style of movement.

Their system sends out a radar pulse and receives the reflected signal. As a person walks, the signal changes because of what is known as the Doppler shift, Greneker said. The signal's frequency increases as it bounces off an object moving toward the biometric system and decreases as the object moves away; the change creates the signature.

In addition, the use of radar means the system can essentially "see" through bulky clothing that might otherwise hide a person's characteristic gait. And by gauging how long the signal takes to return to the biometric system, the technology can precisely measure the distance of a person.

The researchers say their system can recognize a particular person by their walk 80 percent to 95 percent of the time. They hope to improve the accuracy into the high 90-percentile range, but even then, don't believe the technology should stand alone.

"We envision this system as a complementary system to the video-based systems," Greneker said.

The technology isn't likely to move beyond the research center for the next one to five years, according the research institute's Web site.

When it does, however, it could have uses outside of security. Health care providers, for instance, could use the technology to evaluate the changes in a person's gait caused by conditions such as Parkinson's disease and multiple sclerosis.

"We are not sure what the ultimate application will be," Greneker said.

Featured Video

VTech hack exposes 5 million accounts, including kids' photos, chats

The toymaker stores personal data and photos in a way that may be easy for hackers to access. Also, Amazon shows off its latest design for delivery drones.

by Bridget Carey