DCS-Net

This is the implementation of "DCS-Net: Deep Complex Subtractive Neural Network for Monaural Speech Enhancement"

Steps to run the model

1. Edit VOICEBANK_ROOT in config.py to where your copy of the dataset is

2. Tune hyperparameters in config.py

3. Install the relevant modules

$ pip install -r requirements.txt

4. To run DCS-Net:

$ python train.py complex

5. To run DRS-Net:

$ python train.py real

6. To test DCS-Net:

$ python test.py complex

7. To test DRS-Net:

$ python test.py real

Example output files are available in output_files/

Testing DC-Net or DR-Net

In order to test either DC-Net or DR-Net, switch to the speechest branch

File list

data_json/
output_files/
c_network.py
config.py
network_functions.py
r_network.py
README.md
requirements.txt
side_tests.py
test.py
train.py