AquaSignal: An Integrated Framework for Robust Underwater Acoustic Analysis

TL;DR

AquaSignal addresses robust underwater acoustic analysis in noisy, real-world marine environments by integrating preprocessing, deep-learning denoising, supervised vessel-type classification, and unsupervised novelty detection into a single four-stage pipeline. It leverages ORCA-CLEAN (U-Net) for denoising, a ResNet-18 classifier pretrained on ImageNet, and an AutoEncoder for anomaly detection, evaluated on Deepship and Ocean Networks Canada data with a realistic port-monitoring scenario. The work demonstrates 71.3% classification accuracy and 91% novelty-detection accuracy, shows the classifier benefits from denoising yet experiences a modest drop in performance, and provides an extensive ablation study comparing architectures and baselines. These results indicate AquaSignal's potential for real-time maritime monitoring and automated anomaly detection, while highlighting the need for careful data-partitioning, adaptation to other acoustic sources, and future enhancements such as uncertainty estimation and multimodal sensing.

Abstract

This paper presents AquaSignal, a modular and scalable pipeline for preprocessing, denoising, classification, and novelty detection of underwater acoustic signals. Designed to operate effectively in noisy and dynamic marine environments, AquaSignal integrates state-of-the-art deep learning architectures to enhance the reliability and accuracy of acoustic signal analysis. The system is evaluated on a combined dataset from the Deepship and Ocean Networks Canada (ONC) benchmarks, providing a diverse set of real-world underwater scenarios. AquaSignal employs a U-Net architecture for denoising, a ResNet18 convolutional neural network for classifying known acoustic events, and an AutoEncoder-based model for unsupervised detection of novel or anomalous signals. To our knowledge, this is the first comprehensive study to apply and evaluate this combination of techniques on maritime vessel acoustic data. Experimental results show that AquaSignal improves signal clarity and task performance, achieving 71% classification accuracy and 91% accuracy in novelty detection. Despite slightly lower classification performance compared to some state-of-the-art models, differences in data partitioning strategies limit direct comparisons. Overall, AquaSignal demonstrates strong potential for real-time underwater acoustic monitoring in scientific, environmental, and maritime domains.

Figures (7)

• Figure 1: ORCA-CLEAN – Deep denoising network architecture bib2.
• Figure 2: ResNet18 Architecture bib19.
• Figure 3: AutoEncoder Architecture.
• Figure 4: Architecture framework of the proposed pipeline
• Figure 5: Loss plots