Birds Emotion Classification Using AI

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.