Goat-Vocalizations

VOCAPRA is a multidisciplinary project aimed at improving dairy goat management through the analysis of goat vocalizations. This paper focuses on classifying goat vocalizations using machine learning techniques. We utilized the VOCAPRA all dataset, extracting a wide variety of temporal and spectral features. K-means clustering was applied to visualize the feature space, followed by the training and evaluation of two classical classification models—Support Vector Machine (SVM) and Random Forest—alongside a Neural Network model. Our results indicate that all models effectively classify vocalizations, with Neural Network outperforming the others in terms of accuracy. Index Terms—Goat vocalizations, audio pattern recognition, machine learning, feature extraction.