Eagle Ray Shell-Crashing Acoustic Classification Dataset and Models

This dataset contains acoustic recordings of eagle ray shell-crashing feeding behavior, processed acoustic features in MATLAB format, labeled training datasets, and trained machine learning models for classification of feeding events and prey species identification.

Dataset Contents

The deposit includes:

• Raw and processed acoustic recordings of eagle ray feeding events
• MATLAB (.mat) files containing processed acoustic features, labeled datasets, and feature matrices
• Sound field data from controlled acoustic observations
• Training code for model development and retraining on datasets
• Matched filtering code for acoustic event detection from raw signals
• Testing code for deployment and validation on field data
• Trained machine learning models ready for testing and deployment

Prey Species

The dataset covers three primary mollusk prey species:

• Crown Conch (Melongena corona)
• Banded Tulip (Fasciolaria lilium)
• Hard Clam (Mercenaria mercenaria)

Research Methods

This dataset supports marine bioacoustics research employing:

• Acoustic signal processing and feature extraction
• Machine learning classification approaches
• Deep learning architectures for temporal analysis
• Cross-validation strategies for model evaluation

How to Use

This package provides a complete pipeline for acoustic classification and field deployment:

1. Training: Use the training code with provided datasets and .mat files to develop and validate models
2. Signal Processing: Apply matched filtering code for real-time acoustic event detection
3. Field Testing: Deploy testing code on field data to classify feeding events and identify prey species

The .mat files can be loaded in MATLAB or Python (using scipy.io.loadmat or similar libraries). Trained models are included for immediate testing and deployment. Training code allows users to retrain models with modified parameters or additional data. Matched filtering implementation enables robust acoustic event detection on raw field recordings. Feature matrices and labeled datasets enable reproducibility and supplementary analyses.

Applications

This comprehensive package enables:

• End-to-end reproducible marine bioacoustics research
• Model training and refinement with field validation
• Real-time acoustic event detection using matched filtering
• Prey species identification from field recordings
• Deployment of classification models on new acoustic data
• Fully documented and reproducible analysis pipelines
• Educational applications in signal processing and marine biology
• Acoustic monitoring for ecosystem studies

Technical Details

Users should have access to MATLAB or Python with standard scientific computing libraries for working with the data files and trained models. Installation instructions and dependencies are documented in the accompanying code.