Based on "Real Time Speech Enhancement in the Waveform Domain (Interspeech 2020)" Model's code was taken from here https://github.com/facebookresearch/denoiser

1. During trianing I used pytorch lighting and Sacred for logging, you may want to use different logger.
2. It's possible to train Unet as a baseline, just change the model in train_pl.py. Code for Unet is in 'model' folder and it was taken from here together with some other code snippets: https://github.com/haoxiangsnr/Wave-U-Net-for-Speech-Enhancement . Config file contains parameters for Unet.

inference folder contains usage examples with a few enhanced files. Model accepts an audio file in wav format and produces a 'cleaner' veriosn of it with some noise removed.

You can download model weights here: https://drive.google.com/file/d/1Ih8pZ3n4i6VXgwKFYQfMWu3PwCiPtgpG/view?usp=sharing

Training dataset was a collection of different datasets mixed with noise files.

SEGAN 11572 files used. https://pdfs.semanticscholar.org/ed99/08f71d6521a45093ffc0f9365315c1183604.pdf

Clean and noisy files were mixed with random SNR value, additionally white noise was added randomly with different SNR value. For more details check noisy_records_generator folder, it containes files and configs for dataset generation.

1. Russian Open Speech To Text
   subsets used:

• Public_youtube700_val 7311 file
• Buriy_audiobooks_2_val 7850 files
• Asr_calls_2_val 12950 files

2. M-AILABS Speech Dataset (audio books, a few speakers)
   subsets used:

• Ukrainian 31543 files
• Russian 29440 files

1. FSDnoisy18k 9473 files used
2. TAU Urban Acoustic Scenes 9072 files used

Model was trained 400 epochs. I have not checked quality on VCTK or other public datasets and used a subset of generated dataset with 4k files for validation. Scores may look a little bit low because some of the files in "Russian Open Speech To Text" are noisy but they were needed to add speakers diversity.

STOI: 0.941
PESQ: 2.79

An example before and after.