charec-0.2.zip (2017-10-06, 47.9kb)
If you want to peek into the source code, go to the cgit repository.
Draw a single digit number (1-9) on the canvas using mouse.
Charec identifies handwriting characters using Levenshtein distance.
A common view of Levenshtein distance is that it is a measure of string similarity. You have a pair of strings (maybe English words, DNA sequences or whatever) that you want to know how alike they are, so you input them to the edit distance function and get some distance metric. This is pretty common stuff and nothing is wrong with that view.
A more interesting, albeit less common, view about Levenshtein distance is that it's an universal measure of similarity. That is, it can measure the similarity of anything (e.g. images, languages or graphs) as long as these things are encodable as strings. Here is how it typically goes like:
This is essentially how charec works. It takes a set of strokes traced by a mouse cursor, and compare their similarity by inputting them into Levenshtein distance function as strings. For encoding, I used the Schimke-Vielhauer-Dittman technique [link] and it turned out to work surprisingly well.
I hereby place my work "charec" into the public domain. All my copyrights, including related and neighbouring rights, of this work are abandoned.
2019 Fujimoto Seiji <email@example.com>
Back to ceptord.net