v2.1 Release

v2.1 Release Notes

Changes

• Pytorch version of source separation methods are now available
• MNMF and ILRMA are added to numpy / cupy version

v2.0 Release

v2.0 Release Notes

Changes

• The code of FastMNMF1 is refactored.
• FastMNMF2 and AR-FastMNMF2 are newly added. AR-FastMNMF2 is the extension of FastMNMF2 for joint blind source separation and dereverberation.

v1.0 Release

v1.0 Release Note

Multichannel source separation and speech enhancement methods listed below are available.

• Full-rank spatial model based methods

  • FCA
  • MNMF
  • MNMF-DP (MNMF with a DNN-based source model for speech enhancement)

• Jointly-diagonalizable full-rank spatial model based methods

  • FastFCA
  • FastMNMF1
  • FastMNMF1-DP (speech enhancement)

• Rank-1 spatial model based methods

  • ILRMA
  • ILRMA-DP (speech enhancement)