EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Quality-Controlled Generative Augmentation for North Atlantic Right Whale Upcall Detection Using Contour 1D-VAE

Jongmin Ahn iD, Geun-Ho Park iD, Ho-Seuk Bae iD, Donghun Lee iD

Sensors · 23.07.2026 · 26 (15) · 4679

DOI: 10.3390/s26154679 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 23.07.2026. Verfasst von Jongmin Ahn und 3 weiteren Autor:innen. Beteiligte Einrichtung: Agency for Defense Development, Changwon-si 516852, Republic of Korea. Hauptthemen laut MEDLINE: Whales, Vocalization, Animal. Die Arbeit nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Korean Government.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study proposes a Variational AutoEncoder (VAE)-based Quality Controlled (QC) generative augmentation framework for North Atlantic right whale (NARW) upcall detection. Existing generative augmentation methods can generate synthetic samples; however, they do not provide a sample-level criterion for determining whether each generated sample is a positive sample that contributes to improved detector performance or a synthetic outlier that should be removed. This study learns manually extracted upcall frequency contours using a 1D-VAE and evaluates generated contour candidates in a 10-dimensional acoustic morphology feature space. The QC score is computed with respect to the reference distribution of manually extracted real upcall contours, and stochastic acceptance probabilities for borderline samples around the hard threshold are calibrated using same-call manual re-extraction variability. QC-passed contours are converted into detector training spectrograms using smooth amplitude modulation and Gaussian noise injection based on SNR statistics. Using 30,000 acoustic segments from the Kaggle NARW dataset, Original, Denoising Diffusion Probabilistic Models (DDPM), Contour-VAE without QC, and VAE-QC conditions were compared under the same detector and augmentation budget. VAE-QC with α = 0.97 achieved the highest mean AUC of 0.901, outperforming Original training (0.802), DDPM (0.845), and Contour-VAE without QC (0.810). Feature distribution and QC-score analyses further showed that VAE-QC suppresses morphology outlier tails observed in unfiltered generation. These results indicate that the key factor in generative augmentation is the QC process that defines feature boundaries useful for detector learning and selects synthetic positive samples accordingly.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

WhalesVocalization, AnimalAnimalsAutoencoderAcousticsGenerative Artificial IntelligenceAlgorithms

Stichwörter der Autor:innen

North Atlantic right whaleacoustic sensingdata augmentationdiffusion modelvariational autoencoder

Autor:innen und Institutionen

  • Agency for Defense Development, Changwon-si 516852, Republic of Korea.

ORCID-Profile: Jongmin Ahn, Geun-Ho Park, Ho-Seuk Bae, Donghun Lee

Förderung

  • Korean Government

Zitieren

APA

Ahn, J., Park, G. H., Bae, H. S., & Lee, D. (2026). Quality-Controlled Generative Augmentation for North Atlantic Right Whale Upcall Detection Using Contour 1D-VAE. Sensors, 26(15), 4679. https://doi.org/10.3390/s26154679

Vancouver

Ahn J, Park GH, Bae HS, Lee D. Quality-Controlled Generative Augmentation for North Atlantic Right Whale Upcall Detection Using Contour 1D-VAE. Sensors. 2026;26(15):4679. doi:10.3390/s26154679

BibTeX
@article{Ahn2026,
  author = {Ahn, Jongmin and Park, Geun-Ho and Bae, Ho-Seuk and Lee, Donghun},
  title = {Quality-Controlled Generative Augmentation for North Atlantic Right Whale Upcall Detection Using Contour 1D-VAE},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {15},
  pages = {4679},
  doi = {10.3390/s26154679}
}
Kostenlos starten →