Fast algorithm for complex-NMF with application to source separation

In this paper we consider a Nonnegative Matrix Factorization (NMF) model on complex numbers, in particular, we propose a group complex-NMF (cNMF) model that subsumes the phase-consistency complex NMF for the audio Blind Source Separation (aBSS). Using Wirtinger calculus, we propose a gradient-based algorithm to solve cNMF. The algorithm is then further accelerated using a heuristic extrapolation scheme. Numerical results show that the accelerated algorithm has significantly faster convergence.

[1]  Daichi Kitamura,et al.  Consistent independent low-rank matrix analysis for determined blind source separation , 2020, EURASIP Journal on Advances in Signal Processing.

[2]  Xiaodong Li,et al.  Phase Retrieval via Wirtinger Flow: Theory and Algorithms , 2014, IEEE Transactions on Information Theory.

[3]  P. Smaragdis,et al.  Non-negative matrix factorization for polyphonic music transcription , 2003, 2003 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (IEEE Cat. No.03TH8684).

[4]  Nicolas Gillis,et al.  Accelerating Block Coordinate Descent for Nonnegative Tensor Factorization , 2020, ArXiv.

[5]  Nicolas Gillis,et al.  Accelerated Multiplicative Updates and Hierarchical ALS Algorithms for Nonnegative Matrix Factorization , 2011, Neural Computation.

[6]  Roland Badeau,et al.  Complex NMF under phase constraints based on signal modeling: Application to audio source separation , 2016, 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[7]  Nicolas Gillis,et al.  Accelerating Nonnegative Matrix Factorization Algorithms Using Extrapolation , 2018, Neural Computation.

[8]  Roland Badeau,et al.  Phase recovery in NMF for audio source separation: An insightful benchmark , 2015, 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[9]  Hirokazu Kameoka,et al.  Complex NMF: A new sparse representation for acoustic signals , 2009, 2009 IEEE International Conference on Acoustics, Speech and Signal Processing.

[10]  Nicolas Gillis,et al.  Extrapolated Alternating Algorithms for Approximate Canonical Polyadic Decomposition , 2020, ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[11]  Chih-Jen Lin,et al.  Projected Gradient Methods for Nonnegative Matrix Factorization , 2007, Neural Computation.

[12]  Cédric Févotte,et al.  Positive Semidefinite Matrix Factorization: A Link to Phase Retrieval And A Block Gradient Algorithm , 2020, ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[13]  P. Bouboulis,et al.  Wirtinger's Calculus in general Hilbert Spaces , 2010, ArXiv.

[14]  Paul Magron,et al.  Towards Complex Nonnegative Matrix Factorization with the Beta-Divergence , 2018, 2018 16th International Workshop on Acoustic Signal Enhancement (IWAENC).

[15]  Hirokazu Kameoka,et al.  Complex NMF under spectrogram consistency constraints , 2009 .

[16]  Nicolas Gillis,et al.  Blind Audio Source Separation With Minimum-Volume Beta-Divergence NMF , 2019, IEEE Transactions on Signal Processing.

[17]  Stephen A. Vavasis,et al.  On the Complexity of Nonnegative Matrix Factorization , 2007, SIAM J. Optim..

[18]  Noura Hammas.  Traitement du signal et des images , 2018 .

[19]  Jonathan Le Roux,et al.  FAST SIGNAL RECONSTRUCTION FROM MAGNITUDE STFT SPECTROGRAM BASED ON SPECTROGRAM CONSISTENCY , 2010 .

[20]  Ana de Almeida,et al.  Nonnegative Matrix Factorization , 2018 .

[21]  W. Wirtinger.  Zur formalen Theorie der Funktionen von mehr komplexen Veränderlichen , 1927 .

[22]  Philippe Depalle,et al.  Phase constrained complex NMF: Separating overlapping partials in mixtures of harmonic musical sources , 2014, 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).