Researchers at the University of Manchester, in collaboration with scientists at the Universidad Autonoma de Madrid, have developed a biometric artificial intelligence system that can recognize a person from his steps and ways of walking: In particular, the system can recognize the person simply from the moment he presses a special “tile” of pressure on the floor, analyzing 3D and time data.
The results of the research were published in the IEEE Transactions on Pattern Analysis and Machine Intelligence (TRAMI) earlier in the year, indicating that on average, this system was able to successfully identify almost 100% of cases, with an error rate of only 0.7%.
Currently, fingerprints, face recognition and eye scanning are dominate in the biometric security systems. However, the so-called “behavioral” biometrics, such as walking recognition, are also able to identify unique “signatures” as they arise from a person’s behavior and movement. The team of researchers tested these data using a large number of so-called “cheaters” and a small number of users in three different real-world scenarios, at airport checkpoints, workplaces, and a home environment.
“Every person has about 24 different factors and movements while walking, resulting in each individual having a unique way of walking. So the monitoring of these movements can be used as a fingerprint or retinal scan to recognize and identify a person”,
says Omar Kostyja Reyes, of the University of Manchester School of Electrical / Electronic Engineering, who led the survey.
For the purpose of this artificial intelligence system, researchers have created the largest database of walks in history, with about 20,000 walking marks from 127 people. To gather these data, the team used floor sensors and high resolution cameras. It was this dataset (SfootBD), which was used by Kostyja Reyes to develop the advanced computer models needed for the automatic biometric walking identification presented at TRAMI.
As the researchers emphasize, the advantages of this technology are that it is not intrusive (the passerby does not even need to take off his shoes) and could be adapted to identify people with neurological problems, so that it is also suitable for use in health sector.