Decoding Poultry Welfare from Sound-A Machine Learning Framework for Non-Invasive Acoustic Monitoring.

Area of Science:

• Animal Welfare Science
• Bioacoustics
• Artificial Intelligence in Agriculture

Background:

• Precision livestock farming increasingly utilizes non-invasive methods for animal welfare assessment.
• Poultry vocalizations contain crucial information about health, behavior, and stress levels.
• Interpreting complex acoustic signals requires advanced analytical techniques.

Purpose of the Study:

• To develop and evaluate an integrated analytical framework for interpreting chicken vocalizations for welfare assessment.
• To combine signal-level analysis, machine learning, and deep learning for robust classification.
• To validate the framework's performance using diverse, welfare-relevant acoustic datasets.

Main Methods:

• High-fidelity acoustic signal acquisition and preprocessing.
• Extraction of acoustic features, including mel-frequency cepstral coefficients (MFCCs) and spectral descriptors.
• Classification using machine learning (Random Forest, HistGradientBoosting, CatBoost, TabNet) and deep learning (LSTM) models.
• Statistical analysis for feature importance and model validation (cross-validation, F1-score, MCC).

Main Results:

• Specific MFCC bands and spectral features were significantly correlated with poultry welfare indicators.
• Long Short-Term Memory (LSTM) models identified distinct acoustic patterns related to habituation and stressor-specific responses.
• The proposed framework demonstrated high accuracy, generalizability, and interpretability in classifying vocalizations.
• The approach successfully linked acoustic features to known physiological and behavioral processes in poultry.

Conclusions:

• Acoustic sensing coupled with interpretable AI provides a scalable and biologically grounded tool for real-time poultry welfare monitoring.
• This technology supports the advancement of sustainable and ethical livestock production systems.
• The framework's interpretability ensures alignment with established poultry welfare science.