yesterday, i attempted to do OCR from my handwriting on a Tablet. that did not work so well. the results were far from satisfactory. so i stopped going down that path and started down the stroke-based path. the difference being OCR works with images and pixels, and strokes work the x/y coordinates of the pen. beefed up my test app until i could find what looked like a recurring pattern. if i mapped the angle that the pen was traveling to a wave form, then that seemed pretty good. so i shaped that data until the neural net could work with it ... and the pic below shows the very 1st test. 4 characters with no errors ... on the 1st test! ... i was floored :) will have to see what its limits are and then see if i can tweak it to make it better