Exploitable results
Many people pride themselves on never forgetting a face. And rightly so, because recognising and remembering the subtle distinctions that make each look uniquely different is a complex task. Until comparatively recently even the most sophisticated electronic systems were pretty useless at it. But U-FACE has made great strides towards developing an electronic recognition system to control access to restricted information or premises.
Facial recognition has many advantages over current systems, such as passwords or tokens, which can be forgotten, lost or shared. Moreover, it is natural, culturally acceptable and easy to use. In addition, advances in digital photography and video technology mean that facial recognition systems are now becoming an economically viable and practical alternative to other means of authenticating a person's identity.
The U-FACE project, funded by the European Commission's IST programme, has applied and extended existing facial biometrics technology. The system works through a Web cam that scans the face of the person requesting access. The sophisticated software checks the image to see if it tallies with one of the stored 'biometrics templates' of authorised faces. If the face is not 'recognised' by the system, access is denied. As a failsafe, each user must also give a password and has to swipe with a smartcard that contains the digitised data defining his or her face.
Tried and successfully tested
The system has been tried out in two contexts. The first was with a medical practice in Greece. U-FACE was used to allow patients to gain access to a secure website to arrange appointments with their doctor. The second trial was with a publishing company in Italy, where the system was used to enable only authorised personnel to enter the company's offices.
The results of U-FACE were extremely promising, but as project coordinator, Tim Cootes, a senior lecturer at Manchester University in the UK, explains, further work is still needed: "No facial recognition system works really well yet. A lot depends on the quality of the image, and there are still problems with things like make-up, hairstyles, beards and moustaches, spectacles and contact lenses. U-FACE, in common with other systems, has an error rate varying between 5 to 20 per cent, depending on the quality of the face images. The system can be tuned to make access easier (increasing the chance an imposter will be allowed in) or more difficult (increasing the chance a legitimate user will be kept out).
In the biometrics field, no one system is secure enough to use on its own except perhaps iris recognition. But combine two or three measures together, and you have a really powerful way of accurately identifying people.
"In the future, biometric systems will be used in all sorts of contexts, such as immigration control. Often systems don't have to be perfect, as they are used as a deterrent as much as anything else.
"A novel feature of the project was that we investigated how much users liked it. We tried it out on over 50 people and discovered that it was very important to train users and explain how the system worked. As long as you did that, people really liked it. They were both enthusiastic and intrigued."
Another very important feature was the unique statistical models of facial appearance developed by Manchester University in the UK.
Says Cootes: "These models define all sorts of ways a face changes from one person to another - shape, colour, texture and so on. The system has been trained and refined on hundreds of people to collect all possible variations. Using the system we can now create a statistical model of practically anyone and can synthesise an almost infinite number of different faces."
From research to commercial application
This has led to a promising commercial spin off. A new company, Genemation Ltd, has been set up in Manchester with venture capital funding of some 300,000 English pounds to service the computer games industry. The computer games industry currently invests a large amount of time and effort creating realistic faces for use in games. Much of the work is still hand done by artists and results are not always very satisfactory. Genemation will use the techniques developed by the U-FACE team to make virtual face generation not only quicker and far easier, but also more convincing.
Other commercial applications of U-FACE being explored are its use in the financial services industry, the healthcare industry, where secure access to records is paramount, and also in the context of e-learning, where it could be used for authenticating candidates taking online examinations. In the UK, government funding has been granted to continue the work on using the system for gaining access to secure premises, in tandem with a car number plate recognition system.
Source: Based on information from U-FACE
Promoted by the IST Results Service
Buscando datos de OpenAIRE...
Se ha producido un error en la búsqueda de datos de OpenAIRE
No hay resultados disponibles