Handheld Device Reads Printed Words to the Blind

geekotourist writes "3,000 people in Dallas this week for the National Federation of the Blind convention are getting a demonstration of what life is like when you can read printed menus, mail, business cards and memos," reports the Dallas Morning News. The NFB spent two million dollars developing the $3,495 Kurzweil-National Federation of the Blind Reader, which weighs 15 ounces and combines text-to-speech with sophisticated OCR. The device 'gives the user an initial "situation report," describing what it can see. The user then makes a decision about whether to take a picture. After a few seconds to process the image, the contents of the document are read aloud.' Beta testers describe the joys of reading receipts, CDs, food labels, bulletin boards, conference printouts, or of simply reading books with privacy, without another person's help."

Comments

why not braile output?

[DaCool42]

Wouldn't braile output be better? It would allow for more privacy without the need for headphones, and I suspect most blind people could read it faster.

Re:why not braile output?

[Dis*abstraction]

Smartest, wittiest, most stylish guy I ever knew was blind. Got all the ladies. Rhodes scholar, too. And he wrote a screen reader for Linux.

Re:why not braile output?

[jfmiller]

Braile is not necessarly always the best solution. First a large number of vision impaired people never learn it, esp those that go blind late in life. Second while braile can be read as fast as typed print, it takes a good deal of space. This device is too small for a good surface.

As a side note, my supervisor is blind and has a device like this of the desktop varity. He can "read" about 300 words per minute, and be doing other things at the same time. I have fine vision but the though of being able to listen to my textbooks while doing the dishes almost justifies the $2500 price tag.

JFMILLER

Re:why not braile output?

[hooeezit]

This is a common misconception that Braille is the easiest form of presentation for the visually impaired. But that is not so.

Mind you, I use the word 'visually impaired', and not 'blind' for a good reason. A large proportion of the people considered legally blind do have some vision - they fall into the category called 'Low Vision'. There are about 2 million people in the US at present who have Low Vision, but number will swell significantly as the baby boomers age into 50+. Most visual impairments are actually age related, and when you've had vision till age 55 and you suddenly lose it in 6 months, it's a very disturbing experience. Most people who undergo that experience either do not have the ability to or don't care about learning tactile braille at that stage. Even as of now, only a fraction of the visually impaired population can actually read braille.

Also, as the other poster mentions, braille devices are extremely expensive, require a lot of power and are bulky (both in size and weight). A braille display with 40 braille cells will cost an additional $2500.

All that said, I should also mention that building a purely verbal user interface for 'describing' things is a very challenging task. I've been working the last 2 years on a similar device but purely for addressing navigation issues for the visually impaired. We already have a prototype device that can read special barcodes at a distance of about 6 feet, and then that barcode can be looked up in a database to determine the user's location. But how to describe their current location in a manner relevant to their task is proving to be a very tricky problem to solve. Every few months, we feel that we are very close and then discover one more issue that sets us back another few months.

So, it's encouraging to see that someone has been successfully able to build a verbal only interface for descriptive tasks.

- Rudrava Roy
Minnesota Laboratory for Low Vision Research
University of Minnesota, Minneapolis

Re:Inventor is Raymond Kurzweil, Singularity guy

[treeves]

I went to a concert back in 1984 in college where everyone in the audience put on headphones and the performer (I can't remember his name) used a synthetic human head with microphones embedded in it to simulate acoustically the human head (and this was a Kurzweil invention IIRC).
 
  He placed the head inside a grand piano and played - the effect was striking (no pun intended). He tapped and scratched the head and it sounded like he was doing it to my head. What a memory!

Re:why not braille* output?

[robbak]

The comparison is between a complex device made in the dozens, and a complex device made in the billions.

A braile display, which needs to display a line of text - a single changing character wouldn't work, as users slide fingers across the characters - is expensive to produce in the small numbers required.

A sound chip and headphones are used in every mp3 player, HPC and computer in existance. Probably ~50c in bulk amounts.

And as for speed: People who use file readers often have them set to run at 2x-4x speed. As long as the diction is good, it's easy to understand. Especially if you are used to it.

Okay, so the bar has been set...

[Qubit]

How hard would it be to come up with a FOSS system to do the same thing? It sounds like the software makes up a good deal of the cost of the device -- with the proper patrons (like the NFB), perhaps you could come up with some system that would just cost as much as the hardware. I mean, heck, the NFB sunk $2 million into the project, and the blind will still have to pay $3500 for the device.

So you'd start with a good digital camera and a small handheld device. Then you need OCR -> text and text -> speech. What's the state of research or code that one could use in FOSS projects? It's been a year or so since I last checked, but AFAIK the current OCR software that's Free just doesn't stack up with that latest commercial products....

great for blind grad students

[gareth.fletcher]

A cell-phone for the blind was recently made available to visually impaired people in New Zealand, costing around $300USD. It seems like only a small step further to add some sort of camera/document scanner... This particular device will unquestionably help visually impaired students of particular sciences (e.g. advanced math), where there is almost no demand of Braille versions of textbooks (and even the regular textbooks!) and too many books to pay the conversion to Braille (here I believe it's at least $500?).

Portable vs handheld- different

[Anonymous Coward]

You talking about the portable reader- the one that was about the size and weight of a laptop? I saw it a couple of years ago. That isn't the same as software running on a handheld: putting pattern recognition on a Treo sized thing takes some development.

it doesn't look like either of Kurzweil or the National Federation of the Blind are big corporations, or not even a small corporation. And since Kurzweil started working on readers in 1975, I think his dreams could have been big enough to see this coming. It's what all his books are about.

Ray Kurzweil started in 2002, or 1975...

[geekotourist]

In a news article interviewing Ray Kurzweil, it says that he started on the software for the K-NFB reader in 2002: "Kurzweil said the key to being a successful inventor is predicting what the technology will be years from now. That's what he did with this reader. He started developing the software four years ago." Given that he also has a decades long track record in building reading machines, and that other groups have worked on reading machines, the idea that ASU was the first or the only group to be working on this in 2003 isn't entirely plausible.

The first description of this idea - although not as a handheld- seems to have been made in 1934, where ' In his 1934 story The Lost Language, writer David H. Keller describes a device that is actually able to make speech from printed text--the sound-transposing machine.'