EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Sound Source Localization Using Hybrid Convolutional Recurrent Neural Networks in Undesirable Conditions

Bastian Estay Zamorano, Ali Dehghan Firoozabadi, Alessio Brutti, Pablo Adasme, David Zabala-Blanco, Pablo Palacios Játiva, Cesar A. Azurdia-Meza

Electronics · 10.07.2025 · 14 (14) · 2778

DOI: 10.3390/electronics14142778 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Electronics, online erschienen am 10.07.2025. Verfasst von Bastian Estay Zamorano und 6 weiteren Autor:innen. Die Arbeit wurde bisher 6-mal zitiert (Crossref) und nennt 53 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: ANID/FONDECYT Iniciación, cost center, Department of Electricity, Projects: DICYT Regular.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Sound event localization and detection (SELD) is a fundamental task in spatial audio processing that involves identifying both the type and location of sound events in acoustic scenes. Current SELD models often struggle with low signal-to-noise ratios (SNRs) and high reverberation. This article addresses SELD by reformulating direction of arrival (DOA) estimation as a multi-class classification task, leveraging deep convolutional recurrent neural networks (CRNNs). We propose and evaluate two modified architectures: M-DOAnet, an optimized version of DOAnet for localization and tracking, and M-SELDnet, a modified version of SELDnet, which has been designed for joint SELD. Both modified models were rigorously evaluated on the STARSS23 dataset, which comprises 13-class, real-world indoor scenes totaling over 7 h of audio, using spectrograms and acoustic intensity maps from first-order Ambisonics (FOA) signals. M-DOAnet achieved exceptional localization (6.00° DOA error, 72.8% F1-score) and perfect tracking (100% MOTA with zero identity switches). It also demonstrated high computational efficiency, training in 4.5 h (164 s/epoch). In contrast, M-SELDnet delivered strong overall SELD performance (0.32 rad DOA error, 0.75 F1-score, 0.38 error rate, 0.20 SELD score), but with significantly higher resource demands, training in 45 h (1620 s/epoch). Our findings underscore a clear trade-off between model specialization and multifunctionality, providing practical insights for designing SELD systems in real-time and computationally constrained environments.

Förderung

  • ANID/FONDECYT Iniciación
  • cost center
  • Department of Electricity, Projects: DICYT Regular
  • ANID Vinculación Internacional
  • Universidad de Las Américas
  • SENESCYT “Convocatoria abierta 2014-primera fase, Acta CIBAE-023-2014”
  • Pontificia Universidad Católica del Ecuador

Zitieren

APA

Zamorano, B. E., Firoozabadi, A. D., Brutti, A., Adasme, P., Zabala-Blanco, D., Játiva, P. P., & Azurdia-Meza, C. A. (2025). Sound Source Localization Using Hybrid Convolutional Recurrent Neural Networks in Undesirable Conditions. Electronics, 14(14), 2778. https://doi.org/10.3390/electronics14142778

Vancouver

Zamorano BE, Firoozabadi AD, Brutti A, Adasme P, Zabala-Blanco D, Játiva PP, et al. Sound Source Localization Using Hybrid Convolutional Recurrent Neural Networks in Undesirable Conditions. Electronics. 2025;14(14):2778. doi:10.3390/electronics14142778

BibTeX
@article{Zamorano2025,
  author = {Zamorano, Bastian Estay and Firoozabadi, Ali Dehghan and Brutti, Alessio and Adasme, Pablo and Zabala-Blanco, David and Játiva, Pablo Palacios and Azurdia-Meza, Cesar A.},
  title = {Sound Source Localization Using Hybrid Convolutional Recurrent Neural Networks in Undesirable Conditions},
  journal = {Electronics},
  year = {2025},
  volume = {14},
  number = {14},
  pages = {2778},
  doi = {10.3390/electronics14142778}
}
Kostenlos starten →