[podcast]https://edwardwillett.com/wp-content/uploads//2011/09/Gait-Recognition.mp3[/podcast]
Twelve years ago, I started a science column with this sentence: “Are you fed up with having to carry 2,762 separate plastic cards in your wallet for buying gas, getting Air Miles, withdrawing money, renting videos and collecting frequent-ice-cream-eater points? Then you’ll be glad to hear about biometrics…”
More than a decade later, I can’t help but notice that I still have 2,762 separate plastic cards (a rough approximation, admittedly). But work continues on biometrics, and a new study describes a promising new way to use biometrics to pinpoint identity: gait recognition.
Biometrics (as I wrote 12 years ago) “is the measurement of tiny differences among individuals for the purposes of identification. Fingerprinting is probably the best known example.”
Fingerprinting has been around long enough that everyone knows that each person’s fingerprints are unique, which makes their use for identification purposes well-accepted. They also have an advantage in that they’re particularly easy to digitize, and hence to search by computer.
The downside to using fingerprints is that they are closely associated in people’s minds with criminal investigations, making their use in other arenas (such as requiring people who are cashing welfare cheques to place their finger in a scanner, something Connecticut began requiring in 1997) problematic.
There are other forms of biometric identification, of course. There’s the hand scanner, which takes two infrared pictures of a hand, one from above and one from the side, and compares dozens of measurements, including width, thickness and surface area, with a previously stored template. The military has used these for a long time.
Iris scanning is based on variations in shading in an individual’s iris, a pattern that is unique to each individual and also (very important) stable over time. Retinal scans shoot a beam of light into the eyeball and record the formation of veins.
Facial recognition involves photographing faces, then analyzing details of the features and their positioning. Voice recognition measures the wave patterns generated by the tone of a person’s voice.
Now comes gait recognition. According to a study just published in the Journal of the Royal Society Interface in the U.K., footsteps are as unique as fingerprints, and can identify people with a 99.8 percent accuracy.
The method is based on the use of pressure-sensing mats that can record how hard people’s feet press against the ground, their stride length, and other aspects of their gait. These mats are already used in orthotics to help diagnose and treat foot problems.
Gait recognition as a method of identification was tested in a study out of Arizona State University last year. Test subjects walked over a large-area, high-resolution pressure-sensing floor. From that information, a computer model of each individual’s way of walking could be constructed. In that study, the recognition rate (albeit it with only 11 subjects with different walking styles) was 92.3 percent, with a false alarm rate of 6.79 percent.
The small number of test subjects made it difficult to determine if the method could be used in the real world. In the new study, led by Todd Pataky of Shinshu University in Nagano, Japan, 104 people were asked to walk across a half-metre-long board studded with thousands of pressure sensors. Ten steps per person were recorded, with the sensors registering both how each foot applied pressure to the ground, and how the pressure distribution changed as the person walked.
That information was used to train a computer program to pick out the patterns in people’s walks. Out of 1,040 steps recorded, the program incorrectly identified only three.
Pataky told the British magazine New Scientist that, “Even if they have the same foot size, even if they have the same shape, [people] load their feet differently and they do it consistently.”
One application, as I noted earlier, could be in airport security. Each person boarding a plane would walk across a similar sensor (they’re available commercially for about $20,000 each) to ensure they were who they said they were (assuming, of course, that their particular stride was already in the computer database).
The downside: you have to walk across the sensor barefoot, where thousands have walked before. (Yuck.)
But, hey, we’re talking about the airport. In the age of full-body scanners, what’s another humiliation more or less?
