A high-resolution and interpretable sound source localization network with a physics-informed perceptual loss function for enhanced localization accuracy
Measurement Science and Technology · 05.06.2026 · 37 (23) · 236101
DOI: 10.1088/1361-6501/ae70d2 ↗Kurz gefasst
Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 08.06.2026. Verfasst von Yuying Fang und 4 weiteren Autor:innen. Die Arbeit nennt 29 Literaturangaben. Gefördert von: Scientific and Technological Research Program of Chongqing Municipal Education Commission, Natural Science Foundation of Chongqing, China, National Natural Science Foundation of China.
Abstract (Originalsprache)
In the realm of sound source localization, conventional beamforming often confronts the challenge of low resolution. To tackle this issue, numerous enhanced beamforming approaches have been put forward. Nevertheless, these methods still suffer from drawbacks such as suboptimal positioning accuracy and relatively slow calculation speed. In this research, a projection gradient descent method founded on deconvolution approach for the mapping of acoustic sources (DAMASs) was developed and further extended into an interpretable network (DAMAS-PGD-Net), with the aim of achieving precise sound source localization. Within this network, an innovative loss function was meticulously designed. By incorporating weight factors, the energy distribution was optimized, thereby enhancing the accuracy of sound source localization. This method undertakes numerical investigations through the quantitative analysis of the position error, energy error, and imaging time of the localized sound source. Moreover, it has been experimentally verified in acoustic imaging studies. Compared with existing sound source localization methods, the DAMAS-PGD-Net proposed in this paper exhibits superior resolution, a faster calculation speed, and enhanced robustness.
Förderung
- Scientific and Technological Research Program of Chongqing Municipal Education Commission
- Natural Science Foundation of Chongqing, China
- National Natural Science Foundation of China
Zitieren
APA
Fang, Y., Zheng, K., Cheng, Y., Wen, S., & Gou, L. (2026). A high-resolution and interpretable sound source localization network with a physics-informed perceptual loss function for enhanced localization accuracy. Measurement Science and Technology, 37(23), 236101. https://doi.org/10.1088/1361-6501/ae70d2
Vancouver
Fang Y, Zheng K, Cheng Y, Wen S, Gou L. A high-resolution and interpretable sound source localization network with a physics-informed perceptual loss function for enhanced localization accuracy. Measurement Science and Technology. 2026;37(23):236101. doi:10.1088/1361-6501/ae70d2
BibTeX
@article{Fang2026,
author = {Fang, Yuying and Zheng, Kai and Cheng, Yinghui and Wen, Siguo and Gou, Libo},
title = {A high-resolution and interpretable sound source localization network with a physics-informed perceptual loss function for enhanced localization accuracy},
journal = {Measurement Science and Technology},
year = {2026},
volume = {37},
number = {23},
pages = {236101},
doi = {10.1088/1361-6501/ae70d2}
}
EUVIMED