The aim of this research is to develop a method for kansei analysis. The proposed method using Tensor SOM generates three maps, which are the low dimensional visualizations of the evaluators, the evaluation objects, and the evaluation words. Furthermore it is also possible to visualize the relationships between them by coloring the maps. We applied the method to a kansei analysis of landscape images, and it showed us what kind of impression each image has and how the impression varies depending on the evaluators. Finally, we applied this method to a practical task in which we analyzed the instrument panel designs of automobiles.

The map-based direction-giving systems in mobile devices have been widely used, but it is not easy for some users to convert a two dimensional map to their three dimensional surroundings. In this paper, to achieve more intuitive direction-giving, we propose a mobile direction-giving system that employs a human-likely embodied agent depicted in Augmented Reality (AR). We also investigate the effects of two types of gesture use, which come from the different speaker’s perspective (survey and route perspective), on user understanding. We compared the direction-givings when the agent’s gestures were generated in survey perspective and when the agent’s gestures were generated in route perspective. The results showed that participants in survey perspective remembered more than the participants did in route perspective.