Researchers from the University of Manchester and the University of Madrid have teamed up in an AI-driven system that will be able to identify people based on how they walk.
The team trialed their system with 127 individuals who are walking more than 20,000 steps during the process. According to a recent news release, the system was able to successfully identify test subjects more than 99% of the time.
However, it turns out that fooling this system is more difficult than one can imagine. The researchers' AI bases its conclusion on 24 different factors. Manchester's, Dr Omar Costilla, says the results lie in the fact that "every individual person has a unique, singular walking pattern."
The key to this system is that – unlike facial recognition systems – this one does not rely on cameras. Instead, the floor is embedded with sensors that detect footsteps, and that is what makes it so hard to fool the system. Mimicking someone else's gait visually is one thing, but to accurately copy, the force with which they step and the precise distribution of that force, based on the shape of their foot? Just forget about it.
For now, the system is just a concept. Given its accuracy, it would not be too much of a surprise to find that there are airports authorities who want to see how it performs in the real world. It could be a fit for train and bus stations, schools, stadiums and even concert venues.
Companies would use the system to identify employees and visitors. The team even envisions residential applications. The question that remains is whether it is creepier to have cameras scanning your face or that there are sensors hidden in the floor analysing your footsteps.