情報処理学会 第74回全国大会講演要旨

Robots that can understand and express emotion are important for natural human-robot interaction. It has long been suggested that emotion in speech, music and motion have the same underlying form. Based on the literature in these fields, we develop a minimal representation of emotion based on four dynamic features. We show our first steps at evaluating the hypothesized model by developing a system that extracts a description of emotion through speed, intensity, regularity and extent (DESIRE) from human emotional voices, and generates gestures from the same representation. Experiments with the robot NAO show that 1) it is possible to generate emotion-laden robot movements using DESIRE and 2) the same values underlie both speech and movement.