Researchers in Dublin have developed an automated system that proves identity by taking a photo for an ear print, writes Dick Ahlstrom.
A Dublin research group has developed a way to authenticate identity using a picture of a person's ear. The approach has great potential in the growing field of "biometrics", the use of physical attributes to prove identity.
Everyone is familiar with the use of fingerprints or signature for identity purposes, but since the 2001 attacks in New York and Washington and more recently in London, more reliable automated systems are being examined, explains ITT Dublin computer lecturer, James Doody.
He and MSc student Lisa Hickey put together a working prototype of a biometric authentication system based on simple photographs of a person's ear. The ear is underexploited in biometrics research, Doody believes. "The ear seems not to have been used," he says.
A startling number of other attributes have come in for biometric identification, even if the ear hasn't. "There has been a lot of research into face recognition and also hand print and voice recognition," he says.
Researchers have looked at quick measurements of brainwave patters, superficial DNA fingerprinting and thermal images of a person's face. More unusual parameters include the vein pattern on the back of a person's hand and chemical sampling of body odour.
Working systems to authenticate staff on the basis of their walking gait have already been introduced into some US airports, says Doody. These sample aspects of how you walk to confirm you match someone recognised as being a member of staff.
Doody and Hickey took a similar route in her MSc thesis, Authentication from ear curvature. The idea is to authenticate a person's identity rather than produce an identification for an unknown individual.
Doody believes that using a picture of the ear is far superior than say fingerprinting or taking pictures of a person's iris or retina. "It is less intrusive than the eye. The ear is easier to use. You don't want to hold a camera right up to your eye."
Fingerprinting carries a whiff of criminality, he believes, and retinal images can authenticate identity but might inadvertantly also provide personal medical information. "The ear doesn't have any of those issues."
Ear "prints" have been provided as evidence in court cases, but these rely on pressing the ear against a surface. "There is a pressure issue if you don't push it evenly. Ours is different, it is based on a picture of the ear," Doody explains.
A camera captures a two-dimensional image of a person's ear, and image processing is used to create a number of sample points along the outer curve of the ear, its "helix rim". The two created an algorithm that takes measurements between these sample points along the helix rim. "It finds a number of points on the outer curve of the ear and takes slope information. Technically, it is a very simple algorithm," he says.
It produces an identifying template for each person authorised to be on the database and this is compared with fresh images if a person seeks, say, entry to a building. They created a database of staff and students, male and female and of a range of ages to test the system. It delivered 84 per cent correct responses, with about 8 per cent false authorisations and 7 per cent false rejections.
"With more sampling points accuracy would get better. We were trying to keep it simple," says Doody.
He sees many applications of this technology. "Most mobile phones have a camera. You could hold it up to your ear to authorise use of the phone." Similarly it could be used to allow the person gain access to the web or to authenticate an e-commerce purchase, again using the phone's own camera.