Acoustic Based Bird Species Classification Using Deep Learning

This paper uses deep learning techniques to present a novel approach for acoustic-based bird species classification. Leveraging a dataset comprising 10,955 audio samples from six distinct bird species, we extracted diverse features, including chroma, spectrogram, and MFCCs. These features were stacked to form input tensors of size 48x128x1, subsequently fed into a CNN architecture. Our model achieved a commendable accuracy of 0.81 in classifying bird species, demonstrating the efficacy of the proposed methodology. This research contributes to bioacoustics and avian ecology by offering a reliable tool for automated bird species identification, which holds promise for various applications, including biodiversity monitoring and conservation efforts.