Hello!
By, handwriting recognition, you mean English language? If no, then you must have to start collecting dataset for your specific requirements.
Assuming:
We can do CNN (convolution neural nets) for it, to read character by character. It may not be able to recover everything, but like OCR should be able to recover the words.
We have 2 steps: Breaking a sentence into characters (difficult for clumsy handwriting), then identify each character.
Alternatively, if you just wish to identify a person from handwriting, it's not needed, step 1.
Toolset:
MATLAB/Octave (as you've asked) :
Questions?
Whenever you're free, for discussion. I'll need details too.
1) What OTHER algorithms you want co compare to? (Your iv)
2) Do you need any documentation?
3) Any deadlines?
Your iii )
About them, performance of NN can be traced using loss functions like accuracy, soft Max.
About me:
Am a research assistant in machine learning algorithms department. Applications in images (vision), text classification system.
Kindly initiate a chat whenever you're free, and interested. If not online shall reply ASAP.