First, I need to do is to capture my handwriting. That's easy enough as the browser has events for pressure sensitive digital pencils. I've written a little util for capturing data and then another one to convert that data to SVG (like the image above).
The next thing to do is to separate the individual characters. I'm currently leaning towards adding a slicing tool to the capture app and separating the curves manually (like the red lines above). Sure, that could be done with a nerual net, but making one would be slower and the result would probably not be as accurate.