character recognition in Java

Question

I have a new project that invloves on-line character recognition (recognizing characters as they are written). My idea is that each character is defined by a number of strokes that fit a mathmatical model. For instance, 'A' could be defined by three linear strokes; 'O' could be defined as a single circular stroke. The reason for using linear regression would be because not every stroke would be a perfect line or curve every time.Linear regression could be used to form an 'acceptable range' for the strokes.

My questions are:

1. Are there better/easier ways to do this?

2. Does Java have some built in functions or modules that perform linear regression?

3. Are there any languages that could do this easier than Java?

Answer 1

One of the ways to implement this is to build Artificial Neural Network and use associated algorithms to train it. The essence of the method is to build a 2D matrix of 1's and 0's (input matrix). Each element is the input node for the neural network. When we draw a character we set elements we draw over to 1, then you feed the input to your ANN and get actual character as an output. Once you've built a sufficient number of input variations and run the training algorithm over the input sets, you should be able to achieve a much better accuracy then by applying the "linear strokes" method you outlined above.

I think you should be able to find appropriate libraries, which can help you build your ANN and choose appropriate training algorithm. I don't remember any from the top of my head. But you can google around to see what comes up. Cheers!

Answer 2

I found this website where someone wrote and made available their code for several different models of recognizing handwriting. I'm not sure which ones are easier or harder, but I think this will help you. I'm unfortunately unable to answer your other two questions.

Answer 3

My new project idea is to extract a text (word) from an image file. Is that possible? My project submission date is the 1st of March this year.