A ROS-Based Audio Platform for Human-Robot Interaction

Chandra Suryadevara, Ramiz Hussain, Lukas Grasse, Matthew Tata

Keywords: Human-Robot Interaction, Speaker Localization, ROS2, Mobile Robotics, Direction of Arrival (DOA), Voice Activity Detection (VAD), Microphone Array, GCC-PHAT, Turtlebot, Audio Processing, Neuroscience, Autonomous Navigation.

TL;DR: This paper presents an easy-to-integrate, ROS2-based audio system that enables a mobile robot to detect and orient towards a human speaker, achieving success in 80% of trials.

Abstract: Human-robot interaction is facilitated by speech perception and audio awareness, however adding audio systems into robotics faces several challenges. We developed an end-to-end ROS-based solution based on Reverb Robotics audio hardware and software. We built a ROS2 interface and exploited the four microphones of the audio hardware to perform speech activity detection, speaker localization, and speech control of a mobile robot base. This work demonstrates a simple ROS-based way to add audio features to robotics systems that require a high degree of human interaction.

Submission Number: 58