Freakishly accurate A.I. security system identifies you based on your footsteps
From our voice to our faces to more unorthodox methods like our “heartprint,” there are a growing number of ways for computers to recognize their users. One we have yet to come across, however, is based on our footsteps. Researchers from the U.K.’s University of Manchester and Spain’s Universidad Autonoma de Madrid are working to change that. A new A.I. biometric verification method can successfully identify individuals by the way they walk across a pressure pad on the floor. The technology could one day be used as an alternative, or augmentation, to existing biometric identification system in places like airports.
“We have realized that sensing human gait has biometric potential from the early days when we demonstrated the Manchester ‘Magic Carpet’ to user groups back in 2010,” Krikor Ozanyan, Professor of Photonic Sensors & Systems at Manchester, told Digital Trends. “However, decisive progress has been made recently by using artificial intelligence to process the floor senor data in a new way. As the latest development, we teamed up with Universidad Autonoma de Madrid to apply our existing models, and to develop new ones, for experimenting on an existing unique footstep database SFootDB — containing a single stride from a large number of individuals.”
In test conditions, the footstep-recognizing A.I. was able to correctly identify individuals with close to perfect accuracy, with an error rate of just 0.7 percent.
According to Ozanyan, the methodology could be particularly useful for identifying individuals in crowded locations, such as on an airport concourse. Despite a recent story from China claiming that a known criminal was identified using facial recognition in a crowd of thousands, such approaches may not necessarily prove effective. On the other hand, the ability to single out known individuals from a large group of unknowns by getting them to cross a pressure-sensing threshold could be extremely helpful. It could also potentially be used in smart home environments to recognize individual members of the family or household.
“The use of artificial intelligence on footstep data is efficient from the point of view of cost of deployment, resources, upgrades, connectivity and privacy concerns,” Ozanyan continued. “Beyond biometrics, the home scenario is the entry point for the huge area of healthcare, where the occupants’ gait can be monitored over a very long period of time so that subtle changes in their gait patterns can be attributed to normal ageing or to cognitive decline — possibly linked to the early onset of mental illness.”
A paper describing the work was recently published in the journal, IEEE Transactions on Pattern Analysis and Machine Intelligence.