As a society, we’re increasingly comfortable with cameras being a part of everyday life. They’re built into our phones, hanging over every traffic light, and placed behind most cash registers. Still, for law enforcement, a perfect image of a face can be hard to come by, especially when suspects intentionally try to obscure their identity.
It was this issue that Dr. Anil Jain, distinguished professor of computer science and engineering at Michigan State University (MSU), set out to solve Using a database of unconstrained images – also known as “faces in the wild” because pictures are pulled from sources like social media – Dr. Jain and his team (Dr. Dayong Wang, a postdoctoral researcher, and Charles Otto, a doctoral student) created an algorithm that quickly generates a list of candidate matches to help identify unknown faces from surveillance camera footage or crime-scene images.
NEC recently partnered with Dr. Jain and MSU to license this large-scale face-search system and will use it to enhance its current facial recognition solutions.
“NEC has a very powerful face recognition software called NeoFace that was primarily designed for mug shot to mug shot matching,” said Dr. Jain, “and it has performed extremely well compared to its peers in that kind of scenario. So, they were looking for a solution for the problem where query images have rather large variability in terms of pose, illumination, and expression, and still need to be searched against large face databases.”
“NEC is committed to maintaining its leadership position in facial recognition solutions,” said Raffie Beroukhim, vice president, NECAM’s Biometrics Solutions Division. “In addition to our own continued research, partnerships with academia, in particular Michigan State University, is an important aspect of this commitment. We look forward to the fusion of MSU large-scale face-search algorithm with our industry-leading NeoFace facial algorithms to offer more compelling solutions to address ever-increasing security threats and enhance public and national security.”
“What we provided is a prototype,” said Dr. Jain. “NEC will modify the algorithm that we provided, integrate it with their existing systems, and improve the overall face recognition performance.”
Since joining MSU in 1974, Dr. Jain has received numerous recognitions for his contributions to the field of pattern recognition and biometrics, including his February 2016 election to the National Academy of Engineering (NAE), one of the highest professional distinctions bestowed on an engineer.
Interestingly, he didn’t set out to specialize in biometrics. His career took a turn when the U.S. government engaged Dr. Jain, about 25 years back, to find civilian applications for a government-designed hardware, the Splash 2 processor, that was based on FPGA technology.
“They didn’t tell me to work on biometrics, but the hardware that they provided us made us realize that it was extremely suitable for a generic image processing operation, called point matching, where we extract landmarks from two separate images and put them in correspondence or alignment. And since fingerprint matching is done by using point (minutia) correspondence … it was like serendipity.”
For years, biometrics was primarily used for law enforcement and government applications. Over the past five years or so, we’re seeing more consumer applications of biometrics. We use fingerprints to unlock smartphones. There’s even a facial recognition application that can estimate a subject’s age and gender for targeted advertisements. According to Dr. Jain, the rise of biometrics in our everyday lives has had an element of serendipity as well – where market forces have had to align with high usability and low cost to facilitate adoption.
“Who would have imagined just four years ago that everybody would be using a fingerprint to unlock their phones? Biometrics for mobile devices had been available earlier, but it didn’t really become popular until Apple introduced the Touch ID fingerprint sensor in 2013. This shows that sometimes, even though the technology may be ready, the technology doesn’t lift off unless it’s packaged properly – like Apple putting the fingerprint sensor in the home button.”
Dr. Jain and his doctoral students continue to use their research laboratory to investigate real-world issues and address long-standing research problems. In addition to the large-scale face-search system, recent topics include a study on the persistence of fingerprint recognition accuracy over time and methods to prevent printed photo and replay attacks on a face recognition system.
“We’re really proud of the work we did on fingerprint persistence and face spoof detection because these fundamental problems needed to be answered,” said Dr. Jain. “And these are the issues that need to be addressed for every biometric modality. The impact of the problem is what we keep in mind when we choose which topic to work on. Sometimes, more than technology advancement, we take pleasure in advancing fundamental scientific work.”
NEC is proud to collaborate with visionary leaders like Dr. Jain. NEC already has one of the strongest biometrics offerings available, and as we continue our own research and forge partnerships with biometric leaders, the future of biometrics is something to look forward to.