WaveNet-PyTorch

Speech Denoising WaveNet Implmentation in PyTorch

Contents

1. ./main.py or main1.py: python script for training or inference
2. ./data/wavenet/models.py: python script containing speech denoising model
3. ./data/wavenet/denoise.py: python script for speech denoising during inference
4. ./data/wavenet/layers.py or util.py: additional support modules
5. ./data/wavenet/dataset.py: python script for dataset preparation
6. ./data/wavenet/config.py: model parameters
7. ./speech_denoise_test.ipynb: Python notebook for testing model

Dataset

The "Noisy speech database for training speech enhancement algorithms and TTS models" (NSDTSEA) is used for training the model. It is provided by the University of Edinburgh, School of Informatics, Centre for Speech Technology Research (CSTR). Please download from link: https://datashare.is.ed.ac.uk/handle/10283/1942 and save to ./data/NSDTSEA

Original Paper

A Wavenet for Speech Denoising link: https://arxiv.org/abs/1706.07162

Additional files

./data/NSDTSEA/checkpoints contains a pretrained model which can be directly used for inference.