The University of Manchester's School of Electrical and Electronics Engineering, together with the University of Madrid, has developed a behavioral biometric authentication system that can measure a person's personal shape or gait pattern, and enable successful identification of a person simply by walking on a pressure pad on the floor and analyzing the three-dimensional data Dimensional and time based steps.
A model of how to walk including vision, pressure and accelerometers. Illustration: University of Manchester
The University of Manchester's School of Electrical and Electronics Engineering, together with the University of Madrid, has developed a behavioral biometric authentication system that can measure a person's personal shape or gait pattern, and enable successful identification of a person simply by walking on a pressure pad on the floor and analyzing the three-dimensional data Dimensional and time based steps.
Using this system, the researchers claim that the way a person walks and the analysis of that person's steps can be used as biometrics in airport security instead of fingerprints and eye scanning, allowing for a non-invasive method of identity verification.
The results, published earlier this year in the IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) journal that deals with machine learning research, showed that on average the developed artificial intelligence system correctly identified a person nearly 100% of the time, with an error rate of only 0.7%.
The use of physical biometrics such as fingerprints, facial recognition and retinal scans is more common today for security purposes. But behavioral biometrics such as gait recognition can also capture unique signatures derived from a person's natural behavior and movement patterns. The team tested its data by using a large number of so-called "impersonators" and a small number of users in three real-world security scenarios: airport checkpoints, the workplace and the home environment.
Dr. Omar Costia-Reis, from the Manchester School of Electrical and Electronics Engineering who led the project, explains: "Each person has about 24 different factors and movements when he walks, so each person has a unique and unique walking pattern. Therefore, it is possible to use the monitoring of these movements like a fingerprint or retina scan to clearly identify or verify a person."
To create the AI system that computers need to learn such movement patterns, the team collected the largest database of footsteps in history (to date), containing nearly 20,000 footmarks from 127 people. To create the samples and datasets, the team used only floor sensors and high-resolution cameras.
It was the database, called SfootBD, that Dr. Costia-Reis used to develop the advanced computational models necessary for the automatic biometric verification of steps presented at TPAMI.
Dr. Costia-Reis added: "Focusing on the non-invasive detection of the way of walking by monitoring the force exerted on the floor during the step is very challenging. This is because it is very difficult to define manually the differences of subtle variations from person to person. This is why we had to invent an innovative artificial intelligence system to solve this challenge from a new perspective."
One of the important advantages of using step detection is that, unlike taking a photo or scanning at an airport, the process is non-invasive for the person and resistant to environmental noise conditions. The person does not even have to take off his shoes when he walks on the pressure surfaces because it is not based on the form of the step itself but on the way of walking.
Other applications of the technology are smart steps that can detect neurodegeneration, with positive implications in the health sector. This is another area where Dr. Costia-Reis intends to advance his research in step recognition.
He added: "We are also developing the research to address the health issue of markers of cognitive decline and onset of mental illness by using raw step data from wide-area floor sensors that can be installed in smart homes. The movement of people can be a new biomarker for cognitive decline, which can be studied as never before using innovative artificial intelligence systems."
More of the topic in Hayadan:
One response
Although very interesting, it is not clear why *another* biometric method is needed. Face scan/fingerprint is also non-invasive