Environmental Sound Classification Using Data Learning

This project focuses on environmental sound classification using deep learning. The dataset contains audio recordings from 10 different environmental classes, and the task is to train models that can automatically recognize the type of environment or event from a short audio clip.

Dataset Information

Dataset: URBANSOUND8K
Classes: air_conditioner, car_horn, children_playing, dog_bark, drilling, engine_idling, gun_shot, jackhammer, siren, street_music
Features: Mel-spectrograms, MFCCs, log-Mel energy
Labels: Single class per clip

Deep Learning Model Architectures

Selected Model Architectures:

1. RNN (LSTM) + MFCC
2. Custom CNN + LogMel
3. Coordinate Attention Residual Transformer Neural Network + MFCC