Ocular Character Recognition


Ever since individual personal computers first came on-line in large numbers, they have been utilised as a huge opt-in distributed computing array by projects such as SETI at Home and Folding at Home.

But there are information processing tasks that can be distributed yet are still impossible to perform with computers. The Stardust at Home project uses the unparalleled image-recognition capabilities of the human brain to process data from an interplanetary sample collection mission. People all around the world take part in their spare time.

Auntie Beeb’s weekly program on the worldwide use of digital tech, Digital Planet, now reports on another innovative scheme to harness the eyes and noggins of computer users.

You know when a web site displays squiggly text against a blurry background and asks you to type in the characters to prove that you are not a spambot? The CAPTCHA project at Carnegie-Mellon uses this to correct the character recognition of scanned old books.

Optical Character Recognition (OCR) is pretty good these days, partly because the algorithms now use dictionaries and word-frequency tables to improve their guesswork. But the technology is still far more error-prone than a human reader. And OCR can apparently identify which words it’s having problems with. So the CAPTCHA project does something very smart.

When you need to pass a spambot test at a participating website, the project’s server feeds you two words from books it’s working on. One is a word it knows. One is a word it’s having trouble identifying. It doesn’t tell you which is which. You type in both words to gain access to the web site in question. The server thus collects a number of interpretations of each tricky word, and when a certain interpretation gets enough human “votes”, it is accepted as correct. Beautiful! People around the net take hundreds of thousands of these tests every day. Instead of asking people to devote spare time to the project, like Stardust at Home, the CAPTCHA project uses brain time that would otherwise just go to waste.

[More blog entries about , , , ; , , , ]



11 thoughts on “Ocular Character Recognition

  1. Yeah, that’s incredibly neat. But kind of old news; they have been doing that for what, more than a year by now? I’m surprised no major tech news outlets have reported on it earlier. Or have they?


  2. OCR-scanning is the greatest invention since sliced bread. Now if they can just create smarter translating programs my life would be bliss.

    Increadibly smart this CAPTCHA project. I had heard something vague about it but didn’t realize it simply utilised this common security measure. I had started wondering why real words had started to appear instead of just random letters and thought it was meant to make it easier for users. Well, one kan kill two birds with a stone I guess.

    I just hope they don’t get sued for salaries by greedy users.


  3. …maybe it just means it expects the reader to be female? ;-p

    Seriously, I did not know that the CAPTCHA tech was actually doing this. That is damn clever.


  4. Hi Martin..
    I am trying to recognize the alphabet in captcha using template matching ..but still i couldn’t find.
    Could u plz tell me some method to identify the character


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s