This project focuses on developing a machine learning model to identify bird species based on their vocalizations. Given the challenges posed by datasets with imbalanced class distributions, the aim is to curate a selection of bird species that have sufficient audio samples for effective model training. We utilize advanced algorithms, including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and WavNet, alongside comprehensive feature extraction techniques. The audio features extracted include zero-crossing rate, root mean square energy, and Mel-frequency cepstral coefficients (MFCCs), which are pivotal for distinguishing vocal characteristics among species. The model's performance is enhanced through data augmentation strategies, such as noise addition and pitch shifting, to increase the diversity of training samples. This approach allows for robust classification, even with a limited number of audio recordings per species. Ultimately, our model demonstrates the potential to accurately predict bird species based on audio input, contributing to biodiversity studies and ecological monitoring efforts.

Keywords: Bird Sound Recognition, Machine Learning, CNN, LSTM, WavNet, Feature Extraction, Zero-Crossing Rate, Root Mean Square (RMS) Energy,  MFCCs, Data Augmentation, Noise Addition, Pitch Shifting, Species Classification.