Ask HN: Can anyone help me find this swipe/gesture recognition algorithm?
I wonder if someone can help me track down a paper or article I read a long time ago, but haven't been able to find searching google.
The paper was very simple and had some pseudo code that was really easy to implement. I remember it was a really satisfying project to get running but I can't find my old project or the paper now, and want to share it with my brother who is learning to program so he can experience the same joy in getting it working as I did!
The paper described a way of detecting drawn shapes through a simple algorithm that was something like the below:
- User draws a shape in a single 'swipe' on a touch screen.
- You place 10 points along the shape so that they are equidistant across the line they have drawn.
- You adjust the size of those points so they are in a standard 'box' (i.e. make them larger or smaller so that it fits exactly within a standard 100x100 box or similar).
- You compare each of those 10 points against a pre-set list of shapes which each have 10 points (i.e. from a previous recording). Each point on the shape you have drawn is compared against the corresponding point on the pre-set list, and the distance between the points is noted.
- You keep track of which shape has the lowest total distance between points.
- The users shape can be rotated to test if that results in a closer match in a different orientation.
- The shape that has the smallest total distance between all points is the closest-fit gesture and that shape is selected.
Thanks in advance for any help you can give.