A4 Refereed article in a conference publication
A Conceptual Framework for Localization of Active Sound Sources in Manufacturing Environment Based on Artificial Intelligence
Authors: Jalayer, R.; Jalayer, M.; Orsenigo, C.; Vercellis, C.
Editors: Silva, Francisco J. G.; Pereira, António B.; Campilho, Raul D. S. G.
Conference name: International Conference on Flexible Automation and Intelligent Manufacturing
Publisher: Springer Science and Business Media Deutschland GmbH
Publication year: 2024
Journal: Lecture Notes in Mechanical Engineering
Book title : Flexible Automation and Intelligent Manufacturing: Establishing Bridges for More Sustainable Manufacturing Systems: Proceedings of FAIM 2023, June 18–22, 2023, Porto, Portugal, Volume 1: Modern Manufacturing
Journal name in source: Lecture Notes in Mechanical Engineering
First page : 699
Last page: 707
ISBN: 978-3-031-38240-6
eISBN: 978-3-031-38241-3
ISSN: 2195-4356
eISSN: 2195-4364
DOI: https://doi.org/10.1007/978-3-031-38241-3_78
Sound source localization (SSL) is aimed at locating the source of a sound in a space and has been used for decades in many applications, such as robotics, room acoustic analysis, voice communication, and medicine. The main advantages of sound-based methods are their low cost, since they require only a set of microphones, and their high precision in sound source detection due to the possibility of sound penetration through barriers. Although SSL methods have been used in robotics in rescue missions and human-robot interaction, they have not been implemented yet in manufacturing environments, even though the advent of Industry 4.0 and 5.0 manufacturing sectors would benefit greatly from intelligent tools like SSL to make industrial areas smarter. In this paper, a new framework based on SSL is proposed to identify active sound sources like human operators, mobile robots, and machinery in the manufacturing area which can enhance the awareness of a multi-agent system. In our approach, the sound source is estimated through a source region location system based on a Convolutional LSTM method. To make the framework more realistic, a three-stage procedure is proposed, where in the first step only a human and a robot are considered, in the second scenario an asset is added, and in the final stage multiple sound sources are included in the workplace. The proposed framework can improve occupational safety and enhance the cooperation between a human and robot agents in an industrial system.
