TR2024-116

Enhanced Reverberation as Supervision for Unsupervised Speech Separation


    •  Saijo, K., Wichern, G., Germain, F.G., Pan, Z., Le Roux, J., "Enhanced Reverberation as Supervision for Unsupervised Speech Separation", Interspeech, DOI: 10.21437/​Interspeech.2024-1241, September 2024, pp. 607-611.
      BibTeX TR2024-116 PDF Software
      • @inproceedings{Saijo2024sep,
      • author = {Saijo, Kohei and Wichern, Gordon and Germain, François G and Pan, Zexu and Le Roux, Jonathan}},
      • title = {Enhanced Reverberation as Supervision for Unsupervised Speech Separation},
      • booktitle = {Interspeech},
      • year = 2024,
      • pages = {607--611},
      • month = sep,
      • doi = {10.21437/Interspeech.2024-1241},
      • issn = {2958-1796},
      • url = {https://www.merl.com/publications/TR2024-116}
      • }
  • MERL Contacts:
  • Research Areas:

    Artificial Intelligence, Speech & Audio

Abstract:

Reverberation as supervision (RAS) is a framework that allows for training monaural speech separation models from multi- channel mixtures in an unsupervised manner. In RAS, models are trained so that sources predicted from a mixture at an input channel can be mapped to reconstruct a mixture at a target channel. However, stable unsupervised training has so far only been achieved in over-determined source-channel conditions, leaving the key determined case unsolved. This work proposes enhanced RAS (ERAS) for solving this problem. Through qualitative analysis, we found that stable training can be achieved by leveraging the loss term to alleviate the frequency-permutation problem. Separation performance is also boosted by adding a novel loss term where separated signals mapped back to their own input mixture are used as pseudo-targets for the signals separated from other channels and mapped to the same channel. Experimental results demonstrate high stability and performance of ERAS.

 

  • Software & Data Downloads

  • Related Publication

  •  Saijo, K., Wichern, G., Germain, F.G., Pan, Z., Le Roux, J., "Enhanced Reverberation as Supervision for Unsupervised Speech Separation", arXiv, August 2024.
    BibTeX arXiv
    • @article{Saijo2024aug,
    • author = {Saijo, Kohei and Wichern, Gordon and Germain, François G and Pan, Zexu and Le Roux, Jonathan}},
    • title = {Enhanced Reverberation as Supervision for Unsupervised Speech Separation},
    • journal = {arXiv},
    • year = 2024,
    • month = aug,
    • url = {https://arxiv.org/abs/2408.03438}
    • }