Interactive genetic algorithm for font generation system

We suggest the font generation system consists of font definition phase and font creation phase. For font definition phase, the tool to export font element parameters represented by XML format has implemented. For font creation phase, the font creation application PALETTE (Personal Adapted LETTEr) has designed to emerge various fonts based on user's Kansei without hand drawing and developed as Adobe AIR application. It enables to generate a new font by selecting user's favorite fonts from displayed candidates, save font as an image according to users input text, and reuse stored font to the next font creation. To design PALETTE, we focused on Interactive Genetic Algorithm (IGA) technique to take user's subjective desire on font shape into the system. IGA is well known as a method to use user's subjective evaluation interactively. PALETTE has the following modules; (i) create new font candidates base on general font, or stored font characteristics with Kansei label, (ii) generate new fonts using genetic algorithm which can modify each element of the selected candidates, and (iii) export font image and save font characteristics. In this paper, we explain the font generation system and show the experimental results.