Detection and Interpretation of Human Walking Gestures for Human-Robot Interaction

Human-robot interaction (HRI) has become a very important field on mobile robotics community. Service robots must have efficient modules for HRI, in order to deal with humans in a robust way. For a service robot, it is very important to interpret when a person needs or requires assistance. In this paper a system for the detection and analysis of corporal gestures (particularly walking gestures) is presented. The system is used to interpret human attitudes. In order to detect and track humans on interior environments the system uses a PTZ camera and a laser range finder mounted on a mobile robot platform. The digital images obtained through the camera are processed to detect and track people. While a person is tracked by the PTZ camera, pan configuration parameter from camera is used in order to track the legs of the person by the LRF. Motion of the person on the global reference frame is then used to interpret walking gestures.

I. Introduction

In the field of Autonomous Mobile Robots and particularly on Service Robots, significant advances have been achieved in recent years. Nowadays, autonomous mobile robots can construct and maintain maps of dynamic environments in a very robust way [1], [2]. They can plan paths and actions on their environments [3], [4], and they can react to unpredictable situations, as moving objects or persons [5]. With these tasks successfully achieve, scientific teams are moving to new challenges, one of these are Human-Robot Interaction (HRI). There is a lot of work to do, in order to have a really highly interactive robot, beginning from robust people recognition to interpretation of dynamic gestures. The major difficultly on HRI resides on the complexity human's gestures and behaviors.