Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automated data curation for self-supervised learning in underwater acoustic analysis

Hilde I Hummel, Sandjai Bhulai, Burooj Ghani, Rob van der Mei

TL;DR

This paper addresses the challenge of unlabeled, diverse underwater acoustic data by introducing a fully automated data curation pipeline that fuses AIS information with PAM recordings. It employs online hierarchical clustering and AIS-based balancing to construct a diverse, long-tail–aware dataset, enabling effective self-supervised learning via Data2Vec. The curated data improves downstream ship-type classification performance compared with random sampling, though gains vary with environment similarity, highlighting the importance of environmental diversity. The approach paves the way for scalable SSL in underwater acoustics and suggests avenues for expanding dataset diversity and exploring stronger SSL objectives for robust marine monitoring applications.

Abstract

The sustainability of the ocean ecosystem is threatened by increased levels of sound pollution, making monitoring crucial to understand its variability and impact. Passive acoustic monitoring (PAM) systems collect a large amount of underwater sound recordings, but the large volume of data makes manual analysis impossible, creating the need for automation. Although machine learning offers a potential solution, most underwater acoustic recordings are unlabeled. Self-supervised learning models have demonstrated success in learning from large-scale unlabeled data in various domains like computer vision, Natural Language Processing, and audio. However, these models require large, diverse, and balanced datasets for training in order to generalize well. To address this, a fully automated self-supervised data curation pipeline is proposed to create a diverse and balanced dataset from raw PAM data. It integrates Automatic Identification System (AIS) data with recordings from various hydrophones in the U.S. waters. Using hierarchical k-means clustering, the raw audio data is sampled and then combined with AIS samples to create a balanced and diverse dataset. The resulting curated dataset enables the development of self-supervised learning models, facilitating various tasks such as monitoring marine mammals and assessing sound pollution.

Automated data curation for self-supervised learning in underwater acoustic analysis

TL;DR

This paper addresses the challenge of unlabeled, diverse underwater acoustic data by introducing a fully automated data curation pipeline that fuses AIS information with PAM recordings. It employs online hierarchical clustering and AIS-based balancing to construct a diverse, long-tail–aware dataset, enabling effective self-supervised learning via Data2Vec. The curated data improves downstream ship-type classification performance compared with random sampling, though gains vary with environment similarity, highlighting the importance of environmental diversity. The approach paves the way for scalable SSL in underwater acoustics and suggests avenues for expanding dataset diversity and exploring stronger SSL objectives for robust marine monitoring applications.

Abstract

The sustainability of the ocean ecosystem is threatened by increased levels of sound pollution, making monitoring crucial to understand its variability and impact. Passive acoustic monitoring (PAM) systems collect a large amount of underwater sound recordings, but the large volume of data makes manual analysis impossible, creating the need for automation. Although machine learning offers a potential solution, most underwater acoustic recordings are unlabeled. Self-supervised learning models have demonstrated success in learning from large-scale unlabeled data in various domains like computer vision, Natural Language Processing, and audio. However, these models require large, diverse, and balanced datasets for training in order to generalize well. To address this, a fully automated self-supervised data curation pipeline is proposed to create a diverse and balanced dataset from raw PAM data. It integrates Automatic Identification System (AIS) data with recordings from various hydrophones in the U.S. waters. Using hierarchical k-means clustering, the raw audio data is sampled and then combined with AIS samples to create a balanced and diverse dataset. The resulting curated dataset enables the development of self-supervised learning models, facilitating various tasks such as monitoring marine mammals and assessing sound pollution.

Paper Structure

This paper contains 15 sections, 3 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. SSL in underwater acoustics
  4. Data curation methods
  5. Methods
  6. PAM data
  7. AIS data
  8. AIS data curation
  9. Audio preprocessing
  10. PAM data curation
  11. SSL algorithm training and evaluation
  12. Results
  13. Data curation
  14. SSL results curated dataset
  15. Conclusion

Figures (3)

  • Figure 1: The locations of the hydrophones from the web scraped PAM data, indicated in red pointers.
  • Figure 2: The defined range of the selected hydrophones to align raw PAM recordings with the AIS pulses.
  • Figure 3: AIS distribution of individual ships in datset $\mathcal{D}_s$ with threshold value $t=250$ in red, $t=500$ in blue, $t=1000$ in green