Deep Learning-Based Acoustic Classification of Animal Species Using Convolutional Neural Networks

This project focuses on deep learning-based acoustic classification to identify animal species from audio samples. The system employs a combination of machine learning models, including Random Forest (RF), DenseNet + XGBoost, Gated Recurrent Unit (GRU), and Swin Transformer, to classify audio data into 13 distinct animal species categories. Some of the example classes include 'Cat', 'Dog', 'Elephant', 'Dolphin', and 'Monkey'. The features used for classification are extracted using Constant-Q Transform (CQT) and Spectral Contrast techniques, which capture essential frequency and spectral patterns from the audio signals. A user-friendly front-end interface, built using Flask, HTML, CSS, and JavaScript, allows users to register, log in, and upload audio files for classification. After uploading, the system processes the audio and predicts the corresponding animal species based on the trained models. This system offers an efficient and scalable solution for automatic animal sound recognition, with applications in wildlife monitoring, educational tools, and interactive systems.

Keywords: Acoustic classification, deep learning, feature extraction, CQT, Spectral Contrast, Random Forest (RF), DenseNet, XGBoost, GRU, Swin Transformer, animal species recognition, Flask, machine learning.