/DANet

Deep Attractor Network (DANet) for single-channel speech separation

Primary LanguageJupyter NotebookMIT LicenseMIT

Deep Attractor Network (DANet) for single-channel speech separation

This repository provides the implementation of the Deep Attractor Network (DANet) for single-channel speech separation in Jupyter Notebook (.ipynb) format. DANet was introduced in the following papers:

Zhuo Chen, Yi Luo, and Nima Mesgarani, Deep attractor network for single-microphone speaker separation

Yi Luo, Zhuo Chen, and Nima Mesgarani, Speaker-independent speech separation with deep attractor network

Informations about the papers can also be found in our lab website.

Citation

If you find the scripts helpful in your research, please consider citing:

@inproceedings{chen2017deep,
title={Deep attractor network for single-microphone speaker separation},
author={Chen, Zhuo and Luo, Yi and Mesgarani, Nima},
booktitle={Acoustics, Speech and Signal Processing (ICASSP), 2017 IEEE International Conference on},
pages={246--250},
year={2017},
organization={IEEE}
}

@article{luo2018speaker,
title={Speaker-independent speech separation with deep attractor network},
author={Luo, Yi and Chen, Zhuo and Mesgarani, Nima},
journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing},
volume={26},
number={4},
pages={787--796},
year={2018},
publisher={IEEE}
}

Requirements

  • Python 3.6.4
  • Pytorch 0.4.1
  • h5py 2.7.1
  • sklearn 0.19.1
  • numpy 1.15.0
  • librosa 0.6.0
  • jupyter 1.0.0 or above
  • notebook 5.4.0 or above