Gesture encoding and reproduction for human-robot interaction in text-to-gesture systems

Abstract

Purpose - The purpose of this paper is to deal with a method for gesture encoding and reproduction, particularly aiming at a text-to-gesture (TTG) system that enables robotic agents to generate proper gestures automatically and naturally in human-robot interaction. Design/methodology/approach - Reproducing proper gestures, naturally synchronized with speech, is important under the TTG concept. The authors first introduce a gesture model that is effective to abstract and describe a variety of human gestures. Based on the model, a gesture encoding/decoding scheme is proposed to encode observed gestures symbolically and parametrically and to reproduce robot gestures from the codes. In particular, this paper mainly addresses a gesture scheduling method that deals with the alignment and refinement of gestural motions, in order to reproduce robotic gesticulation in a human-like, natural fashion. Findings - The proposed method has been evaluated through a series of questionnaire surveys, and it was found that reproduced gestures by a robotic agent could appeal satisfactorily to human beings. Originality/value - This paper provides a series of algorithms to treat overlapped motions and to refine the timing parameters for the motions, so that robotic agents reproduce human-like, natural gestures