The project 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)
- 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.