[VIDEO] Wearable computer gives visually impaired a way to read

By just pointing her finger at the light in the crosswalk, a computerized voice tells this woman what she is unable to see.

 

The voice is coming from a tiny wearable computer that is attached to the frame of her eyeglasses. It's called OrCam. And when she points at something, the computer is able to understand what information she is looking for, and relays auditory feedback through an earpiece.

 

The pocket size computer may be a game-changer for the nearly 25 million visually impaired people in the United States alone.

 

It enables one to cross busy streets, read newspapers and buy groceries on their own -- something that isn't possible for many with limited eye sight.

 

Hebrew University Computer Science Professor Amnon Shashua, is one of the founders of OrCam.

 

"Visually impaired are people that cannot fix their low vision using optical means, so this could be because of genetic disorders, could be of other ailments that sometimes occur early in life, mostly occur later in life, they see something but not enough in order to manage the daily activities," said Shashua from his office at OrCam's headquarters in Jerusalem.

 

"The purpose of this system is to emulate a helper. Imagine someone standing beside you, looking at the same direction you are looking at understands what you are looking at, understands what information you want to get out of the scene and simply whispers in your ear that particular information."

 

Shashua compared the device to hearing aids, in that it can assist people who can not see but does not treat the ailment.

 

To date, visually impaired people used bulky clip-on magnifying glasses, telescopes of different shapes and sizes and other vision aids or applications on smartphones with limited capabilities.

 

In contrast, the tiny OrCam works for up to six hours before being charged. It is based on a computer vision algorithm called 'Share Boost,' which researcher Yonatan Wexler says has improved artificial intelligence.

 

"One of the hard things about visual perception is the huge amount of information, an algorithm that has to define a bus for example in a picture, has to look through millions of pixels to find that bus so we developed an algorithm called share boost that does this very efficiently and it does it so efficiently that we can fit it in a small device that can fit in your pocket and you can take it with you all day. So the user can stand in a city scape and look around and the device will find the bus among all the pixels in that picture and all the pixels throughout the day that the user is walking around and this is what's unique about our technology," said Wexler.

 

The system, which for now is only programmed to recognize the English language, has very accurate OCR (optical character recognition) capabilities both in terms of speed and response.

 

With limited to sufficient lighting it is capable of recognizing and speaking text on flexible materials such as newspapers, street signs and traffic lights in different fonts, colours and backgrounds with a reasonably high degree of accuracy.

 

Sahshua said there are still some kinks to iron out to improve the OrCam's capabilities, including improving its abilities in low light, as well as its recognition of more languages and handwriting.

 

Currently the device is programmed to identify 500 objects, but is capable of learning an infinite number of new objects which it stores in its memory.

 

Erez Naaman, Vice President of Engineering and Business Development at OrCam, said the system was designed with the importance of comfort and fashion in mind.

 

"We understand that eyeglasses are not just an assistive technology they are also a part of who you are, they are a fashion statement, so you can clip the device onto any kind of eyeglasses just like these and we also made sure to co-operate bone conduction technology to avoid offending (interfering) with peripheral hearing," said Naaman.

 

OrCam will go on sale in the U.S. in September and will retail at $2,500 (USD).

 

Shashua said he hopes to harness computer science to extend the capabilities of the device by adding face and place recognition and allowing people with normal eyesight to navigate using the device combined with GPS (Global Positioning System) capability. — Reuters