Abstract

We propose an approach for a robot to imitate th gestures of ah uman demonstrator. Our framework consists solely of two components: a Sensory-Motor Map (SMM) and a View-Point Transformation (VP ). h SMM establish an association between an arm image and th corresponding joint angles and it is learned byth system during a period of observation of its own gestures.h VP is widely discussed in th psychycT of visual perception and is used to transform th image of th demonstrator's arm to th so-called ego-centric image, as if th robot were observing its own arm. Di#erent structures of th SMM and VP are proposed in accordancewith observations in h uman imitation.h whti system relies on monocular visual information and leads to a parsimonious archTT for learning by imitation. Real-time results are presented and discussed.

Citations