Performance Analysis of “Drive Me” - a Human Robot Interaction System

Abstract

In this paper we introduce a new human-robot interaction system, called “Drive Me” and we make a performance analysis of the system. The system is composed of the human operator, which “drives” the wireless robot using natural arms gestures and the Surveyor's SRV1 robot. The gestures recognition is made using the DTW algorithm and a Kinect sensor. The system is independent of the lighting conditions, it works both in the day light or by night, does not depend of the users' clothes and does not depend of the distance from the robot to the human operator. The robustness of the system is shown by the good gesture recognition rate obtained, which makes of this system a real-time HRI system. We made a performance analysis of the system and we calculated the following performance parameters: accuracy, error rate, precision, sensitivity, specificity, positive and negative likelihood of gesture recognition.