an algorithm or system to recognise hand-written text

Started by
5 comments, last by ProgrammerDX 10 years, 7 months ago

I need a good algorithm that recognizes hand writing and converts it into computer text like OCR for printed text. Any directions?

Advertisement

No idea. And sounds awfully hard honestly.

"I AM ZE EMPRAH OPENGL 3.3 THE CORE, I DEMAND FROM THEE ZE SHADERZ AND MATRIXEZ"

My journals: dustArtemis ECS framework and Making a Terrain Generator

The closest I've seen in my studies towards a goal like this is using neural networks with many, many, MANY test cases. In fact, the first example I had in my neural networks class was a program that could recognize the numbers a person wrote. Perhaps you just try looking for algorithms in that direction...? Honestly I don't know of any other algorithms that might do this...

I wish you luck!!

-ST

EDIT: Here's the link to the class: https://www.coursera.org/course/neuralnets

Look at the first example (video 5 or 6) and they are showing how a neural network can figure out what number a person is writing. Note that this is just numbers....let alone letters, words, etc.

According to http://en.wikipedia.org/wiki/Optical_character_recognition what you want is ICR or IWR and only one bit of software on the list at http://en.wikipedia.org/wiki/Comparison_of_optical_character_recognition_software says it supports that. I suspect with a bit of searching you could find more options.

Systems that can read handwriting do exist in commercial products. For example in automated mail sorting systems.

Just a quick moderation note to say that I edited your title to better reflect the question. Good luck with your problem! smile.png

- Jason Astle-Adams

As was pointed out, this is a hard problem. Hard as in "lots of major companies have invested billions over multiple decades and we still don't have anything that works well."

Just one of the 'easy' cases, a hand-drawn symbol: /-\ Is it an H? Or an A? If it showed up as "H/-\T" or "T/-\E" you could build "HAT" or "THE". If it was standalone it might be "It".
Repeat the process for all variations of handwriting. Then include sloppy handwriting that even the writer struggles to read.

Even high quality commercial OCR and word recognizers fail frequently and require humans to figure it out.
Mail machine processors can detect it... dno if they actually read it tho

This topic is closed to new replies.

Advertisement