Abstract:
The recognition of emotions in animals, particularly birds, through vocalizations holds
significant promise for understanding avian behavior and welfare. In this study, we propose
a novel approach to classify bird species and their associated emotions using artificial
intelligence (AI) techniques applied to vocalizations. Our research focuses on three
commonly encountered species: Parrot, Pigeon, and Budgerigar. We begin by collecting a
diverse dataset comprising vocalizations from various sources, which we subsequently
convert to WAV format for analysis. To tackle the first task of species classification, we
segment audio clips into manageable units and employ feature extraction methods such as
Mel-frequency cepstral coefficients (MFCCs), root mean square energy (RMSE), chroma
features, zero-crossing rate, rolloff, and spectral features. Additionally, we utilize
visualization techniques such as spectrograms to enhance our understanding of frequency
content over time. Employing a range of machine learning algorithms including Gradient
Boosting Classifier, Random Forest Classifier, K-Nearest Neighbors, Decision Trees,
Gaussian Naive Bayes, and XGBoost Classifier, we achieved a fair accuracy in classifying
bird species. Furthermore, we address the challenging task of emotion classification in
birds, focusing specifically on Budgerigars and their vocal expressions of Happiness and
Anger. Utilizing the same AI framework, we have successfully implemented emotion
classification, demonstrating the versatility and robustness of our approach. To the best of
our knowledge, this study represents the first attempt to classify both species and emotions
in birds using vocalizations, thus paving the way for further research in bird emotion
recognition.
