A novel human robot interaction using the Wiimote

Abstract

In this paper, we present a human-robot interaction using Wiimote-based gestures. Our system uses the Nintendo Wii Remote Controller to interact with a real robot. First, the user trains the system using Wii Remote in order to recognize future gestures. For the recognition process, we implemented WEKA Naive Bayes Classifier in C#. The evaluation shows that our algorithm achieves accuracy of 99% with 10 training samples per gesture and 95% with only one training sample per gesture. The novelty of this paper lies in 1) adopting natural gestures compared to traditional perfect geometrical shapes, which are not user friendly, 2) improved accuracy.