Wright wins Lemelson-Illinois Prize for facial recognition work

ECE graduate student John Wright received $30,000 as the winner of the third Lemelson-Illinois Student Prize. The award ceremony was held on March 4 at the National Center for Supercomputing Applications at the University of Illinois. Wright has developed new mathematical tools that drastically improve the accuracy of facial recognition systems.

 “It’s an exciting thing because it’s a great award, and it’s nice in that it recognizes both that the work is nice from a research perspective but also that the potential impact and the fact that we are getting closer to producing a complete solution to robust, reliable, and scalable automatic face recognition,” said Wright.

The Lemelson-Illinois Student Prize is awarded to “an undergraduate or graduate student who has created or improved a product or process, applied a technology in a new way, redesigned a system, or demonstrated remarkable inventiveness in other ways” according the Lemelson Prize Web site.

Wright is currently in Beijing at Microsoft Research Asia (MRSA) and is hard at work applying new mathematical processes to take this face recognition project closer to perfection. “It’s a pretty big topic of research here at MRSA. At the same time, we’d seen a lot of coverage in the popular press” said Wright. “We work on computer vision, but we had never worked on this particular subarea of computer vision before, even though it’s kind of one of the most visible applications of the field.”

There have been attempts in the past to create similar systems, but Wright explained that there have been many high profile failures resulting from technology being deployed in situations that don’t quite meet its operating conditions. In one example in Tampa, Florida, Wright explained that a similar system was tried for a year, but it did not meet with any success. “There were no correct identifications and tons of false ones, so it’s kind of embarrassing to the field,” he said.

Fortunately, new results in applied mathematics may hold the key to make facial recognition possible and successful. Literature discussing sparse representation and compress sensing seemed to fall very easily into the facial recognition problem. “Face recognition under varying illumination with occlusion, falls very naturally into this framework of sparse representation” Wright said. “So it can be cast as a sparse representation problem, and there are new tools to solve that problem.” Wright said that if successful, the area of security is one where this technology could easily be applied.

“Access control is important,” he said, explaining that face recognition could replace key cards for gaining access to buildings or other critical facilities. The technology could also be used for human–computer interfaces, organizing photos, and Internet image searches. “There are literally endless possibilities,” said Wright.

In addition to improving applications with his work, Wright said that it’s simply a very interesting field of study that fits well into his background. “The way these data behave and the methods we apply to them actually tell us methods we can take back that go beyond face recognition,” Wright said. “In a sense its nice both from an intellectual standpoint and from the application standpoint.”

About the Lemelson-Illinois Student Prize
The $30,000 Lemelson-Illinois Student Prize is awarded to a student at the University of Illinois at Urbana-Champaign who has demonstrated remarkable inventiveness and innovation.

The $30,000 Lemelson-Illinois Student Prize is funded through a partnership with the Lemelson-MIT Program, which has awarded the $30,000 Lemelson-MIT Student Prize to outstanding student inventors at MIT since 1995 (see http://web.mit.edu/invent).

For more information on the Lemelson-Illinois Student Prize, please visit http://30kprize.illinois.edu.