Convienient scipt for processing large audio dataset for a time series forecasting experiments.
Features of process_database.py
- Unify sampling rate across whole audio dataset
- Unify data format to
int16 - Unify audio format to
.wav - Split stereo audio files to single channels (kind of data augmentation)
- Rename whole dataset to index based namespace
Additionally split_audio.py can help in splitting every file in dataset for specified time duration audio chunks.
That should help in feeding such dataset in some neural networks.