Autonomous Speech Volume Control for Social Robots in a Noisy Environment Using Deep Reinforcement Learning

This paper presents a novel approach to automatically adjusting the speech volume of a socially assistive humanoid robot to enhance the quality of human-robot interactions. We apply the Deep Q-learning algorithm to enable the robot to adapt to the preferences of a user in the volume of the robot...

Full description

Saved in:

Bibliographic Details
Published in:	2019 IEEE International Conference on Robotics and Biomimetics (ROBIO) pp. 1263 - 1268
Main Authors:	Bui, Ha-Duong, Chong, Nak Young
Format:	Conference Proceeding
Language:	English
Published:	IEEE 01.12.2019
Subjects:	Biomimetics Conferences Control systems Human-robot interaction Humanoid robots Q-learning Reinforcement Learning Social Human-Robot Interaction Speech enhancement Speech Volume Control
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	This paper presents a novel approach to automatically adjusting the speech volume of a socially assistive humanoid robot to enhance the quality of human-robot interactions. We apply the Deep Q-learning algorithm to enable the robot to adapt to the preferences of a user in the volume of the robot's voice in social contexts. Subjective experiments were conducted to verify the validity of the proposed system. Twenty-three human subjects had social conversations with humanoid robots across various noisy environments. Participants rated their perception of the robots' voices in terms of clearness and comfortability through a questionnaire. The results show that the robot equipped with our framework outperforms other experimental robots in trials. This study confirmed the effectiveness of the proposed autonomous speech volume control system for social robots communicating with people in noisy environments.
DOI:	10.1109/ROBIO49542.2019.8961810