Researchers unveil new AI gait recognition system
A team of researchers has developed an artificial intelligence system for biometric gait recognition which can verify an individual’s identity by analyzing the 3D and time data collected by a pressure pad with an error rate of only 0.7 percent, Tech Xplore reports.
Scientists at the University of Manchester and collaborators at the University of Madrid tested the system in airport security checkpoint, workplace, and home environment scenarios, and published the results of research in the journal IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI).
The team collected the largest footstep database created to date, according to the announcement, with nearly 20,000 footstep signals from 127 individuals collected from floor sensors and high-resolution cameras. The SfootBD data set was then used to develop the advanced computational models used for biometric verification when the volunteers and a large number of “imposters” walked across the pressure pads.
“Each human has approximately 24 different factors and movements when walking, resulting in every individual person having a unique, singular walking pattern,” research leader Dr. Omar Costilla Reyes of Manchester’s School of School of Electrical and Electronic Engineering explains. “Therefore monitoring these movements can be used, like a fingerprint or retinal scan, to recognise and clearly identify or verify an individual.”
The technique is non-intrusive and works in noisy conditions without requiring people to remove footwear, as the shape of the footprint is not measured.
“The research is also being developed to address the healthcare problem of markers for cognitive decline and onset of mental illness, by using raw footstep data from a wide-area floor sensor deployable in smart dwellings,” Reyes adds. “Human movement can be a novel biomarker of cognitive decline, which can be explored like never before with novel AI systems.”
The U.S. Defense Information Systems Agency (DISA) has been developing several different types of biometrics for identifying people in the field, including gait recognition.