Start small. Make a program to recognize whether a character is an A or isn't an A. Don't put it on a website, but just on your computer. Keep working on it till it works perfectly, before you even start thinking about other letters than A. Don't give it any fancy input. Just make it always load whatever picture filename you give it, and make it find the character in that picture.

The implementation IS complex, that is because handwriting is complex (from a computers point of view). Get 1000 people to write the letter A and you will have 1000 different looking A's. The computer would have to recognize each one as the same A. Now take the letter l. Is it a small L or a capital i?

Start a project where you have some form of experience. Since you have no knowledge of handwriting recognition, then you shouldn't be starting this project. Develop something you have some knowledge about. If you're good at converting units (like pounds to grams, etc.), then build a web service that does that.

For example, I've recently become interested in handwriting recognition, and want to write something (e.g. an online web app) where the user could hand-write a character with a mouse/tablet, and the app would try to guess and match the correct character. This could eventually have all sorts of applications, e.g. from recognizing English, to Japanese, to maybe even math.

Problem is, I have near zero experience in handwriting recognition. From the existing libraries I've seen, the implementation seems vastly complex and/or requires knowledge a lot of complicated techniques. I have decent coding skills, though, but I haven't really created a big, complex project like this from scratch before.

Is it a wise idea to start working on a project like this, even if I have no idea what I'm doing? Will it be okay if I start it and learn along the way (and refactor later)? If not, how much background research do I need to do before I can actually start the project?