A structured nonnegative matrix factorization for source separation

Clément Laroche (1, 2) , Matthieu Kowalski (3, 2) , Hélène Papadopoulos (4, 2) , Gael Richard (1)

1 LTCI - Laboratoire Traitement et Communication de l'Information
2 L2S - Laboratoire des signaux et systèmes
3 PARIETAL - Modelling brain structure, function and variability based on high-field MRI data
4 labROSA

In this paper, we propose a new unconstrained nonnegative matrix factorization method designed to utilize the multilayer structure of audio signals to improve the quality of the source separation. The tonal layer is sparse in frequency and temporally stable, while the transient layer is composed of short term broadband sounds. Our method has a part well suited for tonal extraction which decomposes the signals in sparse orthogonal components, while the transient part is represented by a regular nonnegative matrix factorization decomposition. Experiments on synthetic and real music data in a source separation context show that such decomposition is suitable for audio signal. Compared with three state-of-the-art har-monic/percussive decomposition algorithms, the proposed method shows competitive performances. Index Terms— nonnegative matrix factorization, projec-tive nonnegative matrix factorization, audio source separation , harmonic/percussive decomposition.

Domaines: Traitement du signal et de l'image [eess.SP]

Clément Laroche, Matthieu Kowalski, Hélène Papadopoulos, Gael Richard. A structured nonnegative matrix factorization for source separation. 23rd European Signal Processing Conference (EUSIPCO 2015), Aug 2015, Nice, France. ⟨hal-01199631⟩