Unofficial Implementation of "Liu, W., Li, A., Wang, X., Yuan, M., Chen, Y., Zheng, C., & Li, X. (2022). A Neural Beamspace-Domain Filter for Real-Time Multi-Channel Speech Enhancement. Symmetry, 14(6), 1081."

• Dependencies
• Data Generation
• Network Training
• Results Computation

Dependencies

• Python, it has been tested with version 3.10.6
• Numpy, tqdm, matplotlib
• torch 1.12.1
• torchaudio 0.12.1

Content

• compute_results.py --> Compute separation results on audio tracks using trained models.
• data_lib.py --> Contains utilities for building pytorch dataset.
• network_lib.py --> Contains network architecture.
• network_modules.py --> Contains modules used to build network architecture.
• train.py --> Trains the model.
• train_lib.py --> Contains utilities for training the networks.