I made a very rough and quick version of how I would do gestures. feel free to give it a try/look.
I suggest clicking once before doing anything. This clears the screen.
Some explanation of that implementation. Basically, I have shapes or gesture symbols created. if the touchpoints follow along the general shape, then a gesture action can be performed.
If I were to take it a step further I would have the gesture shapes try and resize themselves to see if it can find a better gesture match.
I would also have the gesture symbol flip and rotate a few times to attempt and find a better match.
To take the accuracy even further I would place image points on key locations to each gesture and make sure a touch point falls within that area. This would help in accurately determining a gesture.
Im sure there are also other ways to go about this, but this way is quick and mostly effective.