There is more news on ear identification technology (see full article below). New methods for biometric ear identification are being developed by researchers at the University of Southampton. The researchers, using current standards and a database of 63 subjects, have a 99.2% accuracy rate with their new methods. They are also developing new landmarks for measurements. According to a biometrics expert at the University of Southampton, "There are more fixed features available in an ear than we have been measuring." The methods that have been used to compare AA's ear with AN's ear may or may not be superseded by these new methods.
I'll try to explain biometrics, but if this is confusing, please ask and I'll try again. If anyone's interested, I can also provide a bit of general technical background; I'm familiar with some of the algorithms (computer programming math formulas) used in biometrics, but not as they're applied in biometric applications.
There are two basic techniques used for photographic comparisons: anthroscopy and anthropometrics. Anthroscopy is visual assessment and comparing features. Anthropometrics is measurement, and its accuracy depends on measurement tools and methods. Forensic anthropologists use both anthroscopy and anthropometrics to make photographic comparisons.
Biometrics is the use of the computer as a fast and accurate measurement tool for anthropometry. Biometric technologies are used for fingerprint identification, retinal and iris identification, and facial recognition systems.
A summary of the current status of ear identification in biometrics and forensic science:
- the National Training Center for Scientific Support for Crime Investigation in the U.K. is developing an database of several thousand samples. This will be the first statistical proof that ears are unique.
- The first computerized system for ear identification has recently been developed by Professor Guy Rutty, Head of the Forensic Pathology Unit at the University of Leicester, working with a software developer.
- New methods for biometric ear identification are being developed. The article about this follows:
http://www.newscientist.com/article.ns?id=dn7672Ear biometrics may beat face recognition
* 16:09 14 July 2005
* NewScientist.com news service
* Duncan Graham-Rowe
A new type of ear-shape analysis could see ear biometrics surpass face recognition as a way of automatically identifying people, claim the UK researchers developing the system.
The technique could be used to identify people from CCTV footage, or incorporated into cellphones to identify the user, says Mark Nixon, a biometrics expert at the University of Southampton.
Ears are remarkably consistent, he says. Unlike faces, they do not change shape with different expressions or age, and remain fixed in the middle of the side of the head against a predictable background. “Hair is a problem,” Nixon admits. “But that might be solved by using infrared images.”
In an initial small-scale study involving 63 subjects – all taken from a database of face profiles – Nixon and his colleague David Hurley found their method to be 99.2% accurate.
This is a great starting point, says Nixon, but in theory the method could be greatly improved. “There are more fixed features available in an ear than we have been measuring,” he says.
Order of magnitude
Much larger populations are needed to determine how reliably it could be implemented. But an initial analysis of the decidability index – a measure of how similar or dissimilar each of the ears were – indicates how unique an individual ear might be.
They found that this index was an order of magnitude greater than for face analysis, but not as large as for iris biometrics.
Ears have been used to identify people before now, but other methods have used an approach similar to face recognition. This involves extracting key features, such as the position of the nose and eyes – or in the case of the ear, where the channels lie. These are then represented as a vector, describing where features appear in relation to each other.
The new approach instead captures the shape of the ear as a whole and represents this in code, allowing the whole ear shape to be compared.
Ear print
But despite the promising results they may have a job convincing people. In 1998 an ear print left on a window led to the conviction of Mark Dallagher for murdering a 94-year-old woman.
This conviction, however, was overturned in January 2004 because the evidence relating to the ear print was found to be flawed. But Nixon notes that the original evidence was not, strictly speaking, biometric - it relied on a subjective opinion of an ear expert.
The results of their study are due to be published in Computer Vision and Image Understanding.
