Need Support for my Blind Eye Graduation project

Question

Hi all,

i've thought of an idea as a graduation project , but i need some one to support it

i mean, i need to know (just) whether it's Doable Or Not

so that's the idea..

It's a mobile phone application

i want to install the application on A Blind Man Mobile phone , which is supposed to include a powerful camera (say 5 mega pixels)

he should pass the camera on A news paper, story, book Or..Or,...etc and the mobile phone should start pronouncing the text

i know that's OCR (optical character recognition) and image processing needed too,

but i need some one with Experience to tell me Go Or No Go and what fields i should search in... .

if i can contact some one that would be good

Thanks alot...

Answer 1

This is doable, but to make it feasible I would suggest that you strictly outline the performance that you want to achieve initially.

For example, to implement an OCR system for the phone to read printed fonts you could do this by:

1. collected 1000's of examples for each letter that you want to read at a fixed resolution (you can resize) ex. 15x9 pixels
2. collect far more examples of false positives of those letters at same resolution
3. run SVM training on the data set
4. extract for each letter the hyperplane normal that is trained
5. use the hyperplane normal to classify all possible positions in an input image as either being that letter or not, to do this you need to use correlation

Note that the reason this would be feasible on a mobile processor is that the SVM only requires a reasonably cheap computation to classify.