An automatic rule creating method for Kansei data and its application to a font creating system

Abstract

In this paper, we propose a method for creating fuzzy rules of Kansei data automatically. This method consists of 3 steps: (1) Generation of pseudo data of Kansei data by a General Regression Neural Network; (2) Clustering the pseudo data by a Fuzzy ART; (3) Translating each cluster into a fuzzy rule and extracting important rules. In this experiment, we applied this method to "a Japanese font creating system reflecting user's Kansei ." From the result of the experiment, although we have used the same algorithm for drawing font outlines, the system employing our method can reflect Kansei better than the conventional one. © Springer-Verlag Berlin Heidelberg 2005.