- Associated Press - Friday, March 28, 2014

WEST LAFAYETTE, Ind. (AP) - Most drivers don’t consider what their brain is up to as they navigate the road.

Processing the colors of traffic lights, perceiving the movements of neighboring vehicles and discerning the roadway from pedestrians crossing the street are just a few of the slew of tasks being churned out at lightning speed by the human brain.

A Purdue University research team is working to bring that same processing ability to your smartphone.

A research team in the Weldon School of Biomedical Engineering has created the nn-X chip, which would enable cellphones and other devices to identify objects much as the human brain does.


“Right now computers are not exactly smart at understanding the content of images and videos. They see numbers, but they don’t know what’s in the picture,” Eugenio Culurciello, an associate professor at Purdue, told the Journal & Courier (http://on.jconline.com/1pBmWQf ).

Using state-of-the-art algorithms and an artificial neural network, the chip can differentiate objects in the same way a human can easily discern a person from a tree.

A car equipped with that technology could tell which part of the scene is a drivable road and not an out-of-bounds sidewalk - drastically improving driver assistance technology and possibly paving the way for advancements in driverless vehicles.

It could better classify photos or video clips, allowing users to easily sort through their collections on their laptops. A security camera system that records hundreds of hours of video, Culurciello said, could take mere minutes to sort through.

“Wouldn’t it be great,” Culurciello asked, “if the computer could parse the video by itself and then tell you, ‘I saw a couple suspicious people I’ve never seen before and here are their pictures. They were around your house.’ “

For cellphones with such technology, the possibilities are nearly limitless, he said. For example, the phone could use your previous history to help you shop for shoes.

“It could say, ‘Hey, I noticed over there are a pair of red boots you previously looked at and they’re a good price today,’ ” Culurciello said. “It’s physically extending your senses, in a way.”

The idea isn’t entirely new, he said. The capability already exists using large off-site server banks - such as Internet image searches that categorize images by keywords. But such technology has never been placed in a cellphone or put directly in the hands of consumers.

Other teams across the world are working on similar technology, but the speed at which the nn-X chip operates - and its low power consumption - set the chip apart.

“There’s a lot of projects out there that can do what we’re doing, but the big thing we have is we can do the exact same thing with low power,” said Vinayak Gokhale, a graduate student and a member of the research team. “The potential, in my opinion, is very huge.”

Also on the team, which has been working on the project for more than a year, are research associate Berin Martini and graduate students Jonghoon Jin, Aysegul Dundar, Bharadwaj Krishnamurthy and Alfredo Canziani.

Story Continues →