polimi-ispl / rs-mnmf

Ray-Space-Based Multichannel Nonnegative Matrix Factorization

Ray-Space-Based Multichannel Nonnegative Matrix Factorization

About

Matlab implementation of the Ray-Space-Based Multichannel Nonegative Matrix Factorization (RS-MNMF) for audio source separation. A blind source separation is performed adopting the MNMF algorithm to the Ray Space data.

Abstract

Nonnegative matrix factorization (NMF) has been traditionally considered a promising approach for audio source separation. While standard NMF is only suited for single-channel mixtures, extensions to consider multi-channel data have been also proposed. Among the most popular alternatives, multichannel NMF (MNMF) and further derivations based on constrained spatial covariance models have been successfully employed to separate multi-microphone convolutive mixtures. This letter proposes a MNMF extension by considering a mixture model with Ray-Space-transformed signals, where magnitude data successfully encodes source locations as frequency-independent linear patterns. We show that the MNMF algorithm can be seamlessly adapted to consider Ray-Space-transformed data, providing competitive results with recent state-of-the-art MNMF algorithms in a number of configurations using real recordings.

Contents

.
├── LICENSE
├── README.md
├── code
│   ├── lib
│   ├── rayspacenmf.m
│   ├── rsmnmf_example.m
├── data

• code: folder with the source code.
  • lib: folder with utilities for BSS evaluation and more.
  • rayspacenmf.m: MATLAB function for the RS-MNMF.
  • rsmnmf_example.m: example script for RS-MNMF source separation.
• data: folder with the RIR dataset and source signals adopted in the SPL publication.

Usage

Clone or download the repository and run rsmnmf_example.m to see how to use the function rayspacenmf.m.

References

The RS-MNMF for audio source separation was originally proposed in:

• M. Pezzoli, J. J. Carabias-Orti, M. Cobos, F. Antonacci, A. Sarti, "Ray-Space-Based Multichannel Nonnegative Matrix Factorization for Audio Source Separation", IEEE Signal Processing Letters (2021), doi: 10.1109/LSP.2021.3055463

However the following articles are also important for understanding the technique:

• A. Ozerov and C. Févotte, "Multichannel nonnegative matrix factorization in convolutive mixtures for audio source separation," IEEE Transaction on Audio, Speech, and Language, Processing, vol. 18, no. 3, pp. 550–563, 2010.
• S. Lee, S. H. Park and K. Sung, "Beamspace-Domain Multichannel Nonnegative Matrix Factorization for Audio Source Separation," in IEEE Signal Processing Letters, vol. 19, no. 1, pp. 43-46, Jan. 2012.
• L. Bianchi, F. Antonacci, A. Sarti and S. Tubaro, "The Ray Space Transform: A New Framework for Wave Field Processing," in IEEE Transactions on Signal Processing, vol. 64, no. 21, pp. 5696-5706, 1 Nov.1, 2016.

See also

ISPL website, SPAT website