The audio files from the original dataset have been converted to log-mel spectrograms using the following parameters:
Mel frequency bins: 128 (n_mels=128)
Power to dB conversion: Using librosa.power_to_db() with reference to maximum power
Library used: librosa for audio processing
Load audio data from the original Urban Sounds 8K dataset
Extract mel spectrograms using librosa.feature.melspectrogram()
Convert to log scale using… See the full description on the dataset page:
https://huggingface.co/datasets/EthanGLEdwards/urbansounds_melspectrograms.
