Doctoral Student Developing New Facial Recognition System
Source Newsroom: Virginia Tech (Virginia Polytechnic Institute and State University)
Newswise — In biometrics, the facial image is the least accurate means of verifying a person's identification, says Creed Jones, a senior product engineer with Sagem Morpho of Tacoma, Wash., and a Ph.D. engineering student at Virginia Tech.
For Jones, the chair of a national biometrics committee of the American National Standards Institute (ANSI), the watchful eye of Big Brother is a subject that requires common sense guidelines. Photographically, one face may have many different appearances. Problems in verification can result from the angle a picture is taken, or the amount of illumination, or even an expression. To combat some of these shortcomings, Jones is using some very complex mathematics in software programming he is developing.
Jones is working on the development of a face recognition system that uses color. "Everyone currently uses black and white images in biometrics and this presents a problem with skin tone. So I am using the most promising of the black and white technologies, called Gabor filters, and adapting it to color," Jones explains.
The identification of people by their physical and behavioral characteristics is receiving much press in today's heightened state of security. But the early history of biometrics may surprise some. In the 1870s, British soldiers in India had to register their fingertip prints to receive their pay. India also developed the first system for classifying and filing fingerprints in 1897.
By the time the Federal Bureau of Investigation established its central repository of fingerprints in 1924, the science of biometrics was well on its way. Now, into the 21st Century, biometrics is pervasive. Law enforcement agencies are using a proliferation of electronic products to provide recognition of a person's identity, either through eye, voice, face, signature, fingerprint or hand geometry.
"There are a number of societal issues regarding surveillance," says Creed, an electrical engineer. "We will need laws on the use and dissemination of information."
Jones' employer has placed such a high value on his expertise that it has essentially paid his expenses to study for his doctorate at Virginia Tech. When he graduates a year from now, Sagem Morpho also knows that Jones won't be returning to the successful company that garners about one-third to one half of the market share of the biometrics industry. Instead he plans to teach at Seattle Pacific University and stay on his former employer's payroll as a consultant.
Jones selected Virginia Tech for his doctoral studies because of the expertise of Roger Ehrich of computer science and Rich Conners and Lynn Abbott of the Bradley Department of Electrical and Computer Engineering. Abbott is Jones' thesis adviser, and he credits his student with spearheading Virginia Tech's current biometrics-related research.
"Success or failure, as with anything, is in the details," Jones adds. "If we are technically capable of insuring no one can get through a security gate unless the person has plastic surgery," then biometrics technology has climbed to a higher level.
Abbott concurs with Jones' assessment. "The questionable accuracy of today's systems produces too many false alarms, and the facial recognition systems are notorious for not being valid. We need new and better algorithms and theories to focus on the problem, and we need better software.
"We have good cameras and low cost computers, but the algorithms are the hard part. We need to combine statistical analysis with more artificial intelligence analysis."