2009 Volume 8 Issue 4 Pages 1083-1089
In this paper, we propose a font automatic generation system based on collaborative filtering. The proposed system consists of Kansei converter and user interface. Kansei converter receives Kansei words and outputs parameters. In the font generation system, the parameters express the shape of the font. We propose weighted GRNN for the Kansei converter. Weighted GRNN is a new technique that takes in the principle of the collaborative filtering. The system weights a neuron by similarity between users. Kansei converter adds a random number last and outputs multiple candidates. The candidate of multiple fonts is learned in what is chosen by the user. The user interface is interactive. That means that the usage phase and the learning phase are unified. The system can learn the sensitivity of a user with a little burden by this interface. In addition, data gathering is easy for the provider of the system. The following properties about the proposed method are obtained from the experiments. (1) A user can easily make the font that is reflected the sensitivity of the user individual. (2) Data gathering for the system is easy because of the unified user interface. The user evaluation of the proposed system is better than the conventional system.