Beyond traditional wind farm noise characterisation using transfer learning

Phuc D Nguyen; Kristy L Hansen; Bastien Lechat; Branko Zajamsek; Colin Hansen; Peter Catcheside

doi:10.1121/10.0010494

Beyond traditional wind farm noise characterisation using transfer learning

JASA Express Lett. 2022 May;2(5):052801. doi: 10.1121/10.0010494.

Authors

Phuc D Nguyen¹, Kristy L Hansen¹, Bastien Lechat², Branko Zajamsek¹, Colin Hansen³, Peter Catcheside²

Affiliations

¹ College of Science and Engineering, Flinders University, Adelaide, South Australia 5042, Australia.
² Adelaide Institute for Sleep Health, Flinders University, Adelaide, South Australia 5042, Australia.
³ School of Mechanical Engineering, University of Adelaide, Adelaide, South Australia 5005, Australia ducphuc.nguyen@flinders.edu.au, kristy.hansen@flinders.edu.au, bastien.lechat@flinders.edu.au, branko.zajamsek@flinders.edu.au, colin.hansen@adelaide.edu.au, peter.catcheside@flinders.edu.au.

PMID: 36154064
DOI: 10.1121/10.0010494

Abstract

This study proposes an approach for the characterisation and assessment of wind farm noise (WFN), which is based on extraction of acoustic features between 125 and 7500 Hz from a pretrained deep learning model (referred to as deep acoustic features). Using data measured at a variety of locations, this study shows that deep acoustic features can be linked to meaningful characteristics of the noise. This study finds that deep acoustic features can reveal an improved spatial and temporal representation of WFN compared to what is revealed using traditional spectral analysis and overall noise descriptors. These results showed that this approach is promising, and thus it could provide the basis for an improved framework for WFN assessment in the future.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Acoustics*
Machine Learning
Noise* / adverse effects