Towards the Quantification of Human-Robot Imitation Using Wearable Inertial Sensors

Abstract

In this study, we propose a metric in order to quantify how closely a healthy participant imitates a robot, for which we use inertial sensors attached to both individual participant and to a humanoid-robot. For the experiment, twelve healthy participants were invited to perform simple arm movements in order to apply the state space reconstruction which is based on the method of time-delay embedding and PCA. Although the performed arm movements of the healthy participants were very simple, the study reveals that the participants showed different ranges of the proposed metric that can be linked to the level of imitation. Such a metric can be improved in order to determine a detailed scoring of human-robot imitation during training or rehabilitation activities.