CaoYuhang / tcnse

TCN-based Speech Enhancement

TCN-based Speech Enhancement

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 United States License.

This repository is based on funcwj's repository and naplab's repository.

Requirements

Tested in Python 3.7. Please see requirements.txt for other dependensies.

Data preparation

Place the DNS-Challenge folder (or its symbolic link) into data folder. e.g.,

$ cd data
$ ln -s [your path to DNS-Challenge] DNS-Challenge

Then execute make_filelist_testset.sh

$ ./make_filelist_testset.sh

Inference

• configure separate.sh

• run separate.sh

$ ./separate.sh

Paper

If you find this repository useful, please consider citing:

  @article{koyama2020exploring,
  title={Exploring the Best Loss Function for DNN-Based Low-latency Speech Enhancement with Temporal Convolutional Networks},
  author={Koyama, Yuichiro and Vuong, Tyler and Uhlich, Stefan and Raj, Bhiksha},
  journal={arXiv preprint arXiv:2005.11611},
  year={2020}
  }