Iris, facial recognition may obsolete passwords

Article By : R. Colin Johnson

A company has built a suite of algorithms that can eliminate the need for smartphone passwords with face- and iris-recognition.

FotoNation, a subsidiary of US-based Tessera Technologies Inc., claims that its suite of biometrics recognition algorithms can eliminate the need for smartphone passwords with face- and iris-recognition.

FotoNation has long been licensing the "red eye" reduction algorithms used by Nikon and nearly every other camera and smartphone maker making it a no-brainer that they will adopt the new algorithms, according to general manager of FotoNation, Sumat Mehra.

At first their algorithms ran as software on the host application processor, but now the company has created its own hardware accelerator IP that can be licensed to reduce processing time to 100ms and makes its impervious to hackers, according to Mehra. By 2014, its algorithms were being run on 60% of smartphones, and in 2015 it acquired the Mirlin biometric IP for iris recognition, which likewise tracks an individual's facial features.

More recently, FotoNation provided several computational photography and computer vision solutions — HDR, face beautification and panorama technologies — for the OnePlus 3 smartphone, launched June.

"We are already strong in the camera and smartphone markets, but are also planing to bring our new capabilities to the automotive- and surveillance-markets as well," said Mehta. "Mirlin's iris recognition technology has already been proven in European airports, in fighter jets to authenticate pilots and in Iraq to identify people.

[FotoNation Biometrics App cr]
Figure 1: FotoNation’s face recognition solution uses existing front-view/selfie cameras to rapidly identify the subject in various light conditions and poses.

Other algorithms exist for robust iris recognition, but according to FotoNation it has the only IP that does not need connection to cloud computing resources. Running in standalone mode Mehta claims it has a one in 10 million false-acceptance rate, compare to 1-in-10,000 for its closest competitor. Mehra attributes its accuracy to its use of facial tracking of both irises simultaneously.

Its algorithm, as almost all AI algorithms today, is based on a multi-layered neural network performing deep learning on user data. And because it tracks your facial features too, you don't have to stare at the phone to get it to work. It also cannot be "spoofed," according to Mehra, by taking a photo of the user.

Leave a comment