Pointing gesture detection for human-robot communication in informationally structured space

Abstract

Recently, with the improvement of robot technology, human-friendly robots are becoming more familiar to us. The robots should flexibly behave through interaction with human and environments. Human-like-motion has a contribution to provide more natural communication, because the bodily expression can convey important and effective information. Pointing gesture is an important measure to share own cognitive environment with others. Pointing is described as a special gesture functionally in that directing someone's attention to something does not convey a specific meaning in the manner of most conventionalized, symbolic gestures. This paper presents a method of pointing gesture detection for human-robot communication in Informationally Structured Space. Pointing gesture is detected by using posture estimation method based on genetic algorithm. Moreover, pointed object is extracted from point cloud data by using a clustering method based on Growing Neural Gas.