Fish species’ biological vocalizations serve as essential acoustic signatures for passive acoustic monitoring (PAM) and ecological assessments. However, limited availability of high-quality acoustic recordings, particularly for region-specific species like the brown croaker (Miichthys miiuy), hampers data-driven bioacoustic methodology development. In this study, we present a framework for reconstructing brown croaker vocalizations by integrating fk14 wavelet synthesis, PSO-based parameter optimization (with an objective combining correlation and normalized MSE), and deep learning-based validation. Sensitivity analysis using a normalized Bartlett processor identified delay and scale (length) as the most critical parameters, defining valid ranges that maintained waveform similarity above 98%. The reconstructed signals matched measured calls in both time and frequency domains, replicating single-pulse morphology, inter-pulse interval (IPI) distributions, and energy spectral density. Validation with a ResNet-18-based Siamese network produced near-unity cosine similarity (~0.9996) between measured and reconstructed signals. Statistical analyses (95% confidence intervals; residual errors) confirmed faithful preservation of SPL values and minor, biologically plausible IPI variations. Under noisy conditions, similarity decreased as SNR dropped, indicating that environmental noise affects reconstruction fidelity. These results demonstrate that the proposed framework can reliably generate acoustically realistic and morphologically consistent fish vocalizations, even under data-limited scenarios. The methodology holds promise for dataset augmentation, PAM applications, and species-specific call simulation. Future work will extend this framework by using reconstructed signals to train generative models (e.g., GANs, WaveNet), enabling scalable synthesis and supporting real-time adaptive modeling in field monitoring.

This study proposes a TDOA-MLP method for estimating the range of snapping shrimp sources from the array-wide time-difference-of-arrival (TDOA) structure measured by a vertical line array. BELLHOP arrival-time replicas were generated for the SAVEX15 environment by varying source range, array tilt, and maximum water depth, and a multilayer perceptron was trained to estimate source range with array tilt as an auxiliary output. To provide a more stringent assessment of range generalization, six continuous source-range intervals were excluded from training, and only their inner portions were used for blocked-holdout evaluation, providing an approximately 2-m separation from the nearest training ranges. The resulting overall range-estimation mean absolute error (MAE) and root mean square error (RMSE) were 0.864 and 1.056 m, respectively, with regional MAEs of 0.762, 0.728, and 1.119 m for the near-, mid-, and far-range regions. Additional analyses showed that the estimation error increased with controlled arrival-time perturbations and varied with the maximum-depth condition. When applied to experimental SAVEX15 snapping shrimp data, the TDOA-MLP produced range distributions and snap-to-snap variations generally consistent with those obtained using template matching, steered frequency-wavenumber analysis, and frequency-difference matched-field processing. Relative to template matching, the mean absolute relative range difference was 4.54%. Because ground-truth source ranges were unavailable for the experimental data, the field results demonstrate consistency among the methods rather than absolute localization accuracy. The proposed method provides a low-dimensional learning-based approach for estimating the range of impulsive snapping shrimp signals from array-wide TDOA structures.

Passive acoustic monitoring (PAM) is a widely adopted methodology for the continuous, long-term observation of underwater marine organisms as well as various underwater noise sources using acoustic signals. However, manually identifying specific signals within long-term monitoring datasets is cumbersome and susceptible to potential biases. This study proposes a hybrid automated detection framework that simultaneously improves detection accuracy and operational efficiency for continuous 45-day underwater acoustic data on the grunt signals of Pacific cod (Gadus macrocephalus). The framework employs a theoretical kernel matching the dominant signal harmonic frequency components to rapidly filter potential candidates with minimal false negatives, followed by a class-weighted ResNet-18 to eliminate residual false positives arising from severe class imbalance in real-sea environments. To prevent data snooping, the classification probability threshold was systematically investigated on an independent validation dataset; freezing the canonical threshold at 0.5 effectively suppressed false alarms caused by ambient soundscapes. On the chronologically distinct test dataset, the proposed model achieved an outstanding and balanced classification performance, with 97.67% precision, 94.83% recall, and a 96.23% F1-score. These findings demonstrate that the integrated matched filter and deep learning pipeline provides a reliable automated approach for identifying biological fish sounds within complex marine acoustic environments.

Effective identification of shipping signals in underwater environments is essential for maritime operations and ecosystem monitoring. Traditional models require extensive data for each ship type, posing a significant challenge owing to the difficulty of collecting diverse signals, particularly for vessels with security constraints. Few-shot learning offers a promising solution by enabling ships identification from minimal data through accurate template matching. This study proposes a novel few-shot learning approach that leverages stochastic information within and between ship types to improve identification accuracy using limited labeled data. The proposed model is designed based on a Siamese prototype network that integrates intra- and inter-category dissimilarities, employing cosine distance to estimate similarity while accounting for variance within the data. It achieves robust performance even when trained on limited samples with an average accuracy of 87.81% in five-way identification. In addition, its ability to generalize to unseen ship classes highlights its potential for real-time marine applications, further confirming the effectiveness of few-shot learning in constrained data scenarios. This approach provides valuable insights into designing adaptive, efficient systems for underwater signal detection and has potential applications across a wide range of acoustic processing tasks.