Interactive Intention Prediction Model for Humanoid Robot Based on Visual Features

Abstract

It is necessary for humanoid robots to have the capability of human-like social environment perception, which could enable robots to interact with people intelligently. In order to improve the interaction capability of humanoid robots in complex and changeable social environments, an interactive intention prediction model(IIPM) is proposed, which can quantitatively predict the intensity of interactive intention by the visual features of face orientation, social distance and facial expression in the actual social environment. Based on this model, humanoid robots can make autonomous decisions and select interactive person to carry out interactive tasks reasonably. Finally, the prediction accuracy of the IIPM is proved by single-person and multi-person experiments, which provides an effective and accurate solution for natural human-robot interaction (HRI).