Electronic Cocktail Napkin

The Cocktail Napkin uses a simple recognition algorithm to identify multi-stroke hand drawn glyphs. It compares pen path, number of strokes and corners, size, aspect ratio, and rotation of the input glyph with a library of templates. Pen path is identified through the squares of a 3x3 grid, numbered 1-9. 90 degree rotations, reflections, and inversions are obtained by permuting the numbering of the grid. Corners are identified by examining the pen input data and finding local minima in the pen speed.

Features of a hand drawn glyph: 3x3 grid, corners, raw input points.

No special "training mode" is used. As you use the program, you can identify glyphs that the program does not recognize. For example, to teach the program a new glyph, you just draw 10-20 examples on the screen, and enter a name. You can also set a switch to have the program query you whenever it does not recognize what you drew. We prefer to leave this switch off, and allow the program to simply carry unrecognized glyphs.

Various pen paths trained for a "Box"; red circle shows starting point.