A team at the Georgia Institute of Technology said early research into gait recognition suggests the human walk contains unique characteristics which can be defined by analysis of radar signals.
The project’s ultimate goal is to allow gait recognition to join other emerging biometric technologies such as facial and iris identification for use in security systems.
Most similar research has concentrated on the use of machine vision technology, enabling a computer to ‘see’ human movement and then analyse it.
The Georgia team decided to explore the potential of radar – which produces a unique ‘signature’ from any object it scans – as a source of raw data.
By collecting a radar signature from the movement of a human body’s various parts as the person walked, the researchers hoped to define a set of unique characteristics for that individual.
‘We didn’t really know whether we would find anything useful, but fairly soon we began to come out with some distinct features,’ said research engineer Jon Geisheimer.
The team was able to correctly identify between 80-95% of a test group from their radar signature.
‘We have also been able to detect asymmetries such as a limp,’ said Geisheimer, who suggested the technology may eventually have medical as well as security applications.
He also claimed that it was ‘pretty readily apparent’ if someone was deliberately walking strangely in a bid to baffle the system.
Despite their promising early hit rate, Geisheimer admitted gait recognition still has significant hurdles to overcome before it is ready for commercial use.
The initial testing was carried out at short range, indoors and using co-operative subjects.
The Georgia team will now build a more powerful radar system that can detect people at more than 152m, and attempt to develop more robust processing techniques for the data it collects.
It will also test the technology outdoors, where it believes the system could compete with vision-based technology.
‘Unlike a camera, radar can work at night and in all weather conditions,’ said Geisheimer.