Skip to main content
Media appearance

Drucksensoren in der Biometrie: Wenn der Gang die Identität verrät

Deutschlandfunk’s July 2018 feature on footstep biometrics described University of Manchester research on recognizing people from floor-pressure data. It quoted Dr. Omar Costilla Reyes.

Architecture figure for the footstep-recognition paper: each stream passes through convolution blocks with skip connections, a fully connected layer, and a linear support vector machine before score fusion. Shared when the paper was published in IEEE TPAMI, February 2018. Related photograph from Analysis of spatio-temporal representations for robust footstep recognition with deep residual neural networks. Credit: Image courtesy of Dr. Omar Costilla Reyes.

A radio feature on footstep biometrics

Published on July 2, 2018, in the Deutschlandfunk programme Forschung aktuell, the report by Nina Rink, titled “Drucksensoren in der Biometrie: Wenn der Gang die Identität verrät” (Pressure sensors in biometrics: when your gait reveals who you are), describes research from the University of Manchester and presents Dr. Omar Costilla Reyes as an expert in deep machine learning and sensor systems. Dr. Costilla Reyes noted at the time that the research on AI and footsteps as a biometric had been featured on German national radio, and shared the piece with his Manchester colleagues. The coverage arrived in the weeks after his PhD was awarded and as he began a postdoctoral fellowship at MIT. It followed University of Manchester news articles in May and June 2018 on his footstep-recognition research, part of a wave of media attention that reached many outlets.

How a floor can identify a walker

A biometric is a measurable characteristic used to recognize a person. Fingerprints, irises, and faces are physical biometrics; the way someone walks is a behavioral one. A floor sensor records the pressure pattern of each footstep, and a machine-learning model learns which patterns belong to which person. The feature reports that the team developed the system over several years with biometrics researchers from Madrid, had a large group of volunteers walk repeatedly over a pressure plate, and assembled what it calls the largest footstep database so far. It describes airport security checks as one possible use, and quotes Dr. Costilla Reyes on a second research interest: whether changes in movement could help detect neurodegenerative conditions such as Alzheimer’s disease earlier.

Results, and the paper behind them

The feature relays the researchers’ finding that slight changes such as walking barefoot did not defeat the system in their tests, which shows that the footstep signature holds up under everyday variation. The research was published in 2018 in IEEE Transactions on Pattern Analysis and Machine Intelligence, developed with collaborators at the University of Manchester and the Universidad Autónoma de Madrid. That paper reported a very low equal error rate for verifying people from floor-sensor footstep data (the point at which wrongly accepted and wrongly rejected people are equally frequent), and examined three deployment scenarios with different amounts of training data (airports, workplaces, and homes), which shows how the method carries over to the different places where a floor sensor could be installed. This identification work sits alongside the wider gait research described on related pages of this site.

Topics