Gesture Recognition for Human and Robot Interaction Underwater

Abstract

Gesture recognition is a convenient human-machine interaction method and computer vision-based techniques are typically employed for this purpose. In literature, there are many algorithms employed with various degrees of success. However, recent solutions are built upon deep-learning techniques since they provide superior performance in terms of recognition rate and reliability. In this paper, we aim to develop a gesture-based human robot interaction solution particularly applied to underwater robotics. Underwater is a challenging environment for humans as well as robots where wireless communication is severely limited due to the physical properties of the environment. We used a set of hand and body gestures to communicate with an autonomous underwater vehicle (AUV) which is developed to assist human divers (recreational or commercial). Using on board camera and computer system of the AUV and well-known machine learning algorithms, we investigate using gesture recognition as a means to instruct AUV to perform certain tasks. In the following, we will present the machine learning algorithms used and the performance of gesture recognition in underwater environment.