Image processing question - beginning a personal project

Hello,

I'm about to begin a personal project that will deal with some image processing, and I was hoping for advice before jumping in. My goal is to create a program that accepts an image of my hand-written homeworks, and processes each character into, say, a text file (further manipulations will be simple). My biggest problem lies in acquiring the characters from the image.

Let me first tell you my ideas:

1) Provide samples of my handwriting of each character (numbers, letters) and use a simple hamming distance (might have to scale for comparison).

2) Train a neural network to recognize my characters and use that (which NN would you recommend?).

3) Anything else you can think of that's better.

Otherwise I plan to use bounding boxes that I think MATLAB can automatically determine for each character (I assume given some threshold).

My biggest problem; however, within this character determination is that I will sometimes write a fraction,

a
_

b

and I'm not sure if I can somehow detect it's a fraction rather than 3 lines, or if i should stick to a/b.

Question: what method of character recognition would you recommend I employ, given that I will be taking from a hand-written (possibly tilted) input image, and how should I handle multi-line fractions (or anything that has a similar multi-line issue)?

Thanks for your time community.

Edit: although I might have added Neural Network Toolbox, I don't have nor plan to use that - I'll code my own of whatever is suggested, so please don't advise anything directly from that toolbox :).

[José-Luis]

There's an open-source project doing ocr for mathematical formulas.

http://cat.prhlt.upv.es/mer/

https://github.com/falvaro/seshat

So you can either try to implement it in Matlab or contribute to the open-source effort.

Personally, I'd pick the later.

[Greg Heath]

Why don't you use the forward slash for fractions?

Greg

[MATLAB Newbie]

@Jose-Luis That is incredibly cool, and it seems to work decently well - I will try to work with it. Thank you!

@Greg Heath I was considering resorting to that, but when I'm actually writing my homework, I usually see things more clearly in the other format, especially when dealing with lengthy numerator/denominators (+seeing cancellations).

Thank you both for your responses!