Construction and Evaluation of a Virtual KANEI System for Robots

Abstract

In this paper, three components of virtual KANSEI for robots; emotion identifier, emotion generator and action modulator, are integrated to construct a virtual KANSEI system. The emotion identifier detects robot partner's emotion from facial images. The emotion generator, also called the virtual KANSEI in a narrow sense (VKNS), generates time-series of robot emotion. The action modulator modifies robot motion according to the generated emotion. Those three components are integrated to form a virtual KANSEI system, also called the virtual KANSEI of the wide sense (VKWS). The system is implemented on a virtual robot and is evaluated from ‘naturalness’ point of view of the generated robot action.