I really like your design but one thing springs to mind when I look at your screencaps. When I'm learning how to write a complicated kanji, I'd greatly appreciate having as much room on the screen as possible. It affords greater precision and the details are better shown. So maybe make the drawing area bigger?
Now having tried it for just a minute or two, I'm not sure I entirely like your recognition algorithm. It said I was wrong! Seriously though, while my hand writing is pretty sloppy at times, I'd still like to think my hand writing is legible. The algorithm seems a lot less sensitive about straight lines than some of the more curly bits. Take あ for instance. I while the second stroke was well off it accepted that while rejecting the third stroke for not bending back enough way at the end.
Since algorithmic solution is rather hard I might actually suggest going a whole different direction. If you've ever tried applications like Anki, they ask the user herself to select the result, which to me seems to work fairly well. The least I would want is probably some kind of intermediate level in between Good and Wrong.
Hope it helps and good luck!