Deep learning-based sound source separation for quantitative evaluation of the influence of disturbance sounds on aircraft noise
The Journal of the Acoustical Society of America · 01.10.2025 · 158 (4_Supplement) · A390-A390
DOI: 10.1121/10.0041260 ↗Kurz gefasst
Fachartikel in der Zeitschrift The Journal of the Acoustical Society of America, erschienen am 01.10.2025. Verfasst von Tsumugi Nakayama und 5 weiteren Autor:innen.
Abstract (Originalsprache)
To properly evaluate aircraft noise exposure, many aircraft noise events containing other environmental noise (disturbance sounds) are generally excluded, as their signal-to-noise ratio (SNR) cannot be quantitatively measured. If the SNR between aircraft noise and disturbance sounds can be estimated and the contribution of the disturbance sounds is sufficiently small, it is possible to determine whether the data with disturbance sounds can be used for evaluation. This study investigated a method for quantitatively evaluating the influence of disturbance sounds on aircraft noise, using the SNR calculated from the two signals separated by deep learning. Source separation was performed on mixtures combining aircraft noise and disturbance sounds with different SNRs. We used several types of disturbance sounds with different temporal and spectral characteristics. The separation performance was evaluated based on the difference in the A-weighted sound exposure level between the original and the separated signals. Results showed that high separation performance was achieved for impulsive disturbance sounds. Conversely, the performance slightly decreased for disturbance sounds with characteristics similar to aircraft noise, such as long duration and broadband spectral characteristics. The SNRs estimated from the separated signals were generally consistent with the original SNRs.
Zitieren
APA
Nakayama, T., Sato, R., Hiruma, N., Ezoe, T., Yokota, T., & Kouda, S. (2025). Deep learning-based sound source separation for quantitative evaluation of the influence of disturbance sounds on aircraft noise. The Journal of the Acoustical Society of America, 158(4_Supplement), A390-A390. https://doi.org/10.1121/10.0041260
Vancouver
Nakayama T, Sato R, Hiruma N, Ezoe T, Yokota T, Kouda S. Deep learning-based sound source separation for quantitative evaluation of the influence of disturbance sounds on aircraft noise. The Journal of the Acoustical Society of America. 2025;158(4_Supplement):A390-A390. doi:10.1121/10.0041260
BibTeX
@article{Nakayama2025,
author = {Nakayama, Tsumugi and Sato, Ryo and Hiruma, Nobuhiko and Ezoe, Taisuke and Yokota, Takatoshi and Kouda, Shunsuke},
title = {Deep learning-based sound source separation for quantitative evaluation of the influence of disturbance sounds on aircraft noise},
journal = {The Journal of the Acoustical Society of America},
year = {2025},
volume = {158},
number = {4_Supplement},
pages = {A390--A390},
doi = {10.1121/10.0041260}
}
EUVIMED