Multichannel harmonic and percussive component separation by joint modeling of spatial and spectral continuity

Ngoc Duong (1) , Hideyuki Tachibana (2) , Emmanuel Vincent (1) , Nobutaka Ono (2) , Rémi Gribonval (1) , Shigeki Sagayama (2)

1 METISS - Speech and sound data modeling and processing
2 UTokyo - The University of Tokyo

This paper considers the blind separation of the harmonic and percussive components of multichannel music signals. We model the contribution of each source to all mixture channels in the time-frequency domain via a spatial covariance matrix, which encodes its spatial characteristics, and a scalar spectral variance, which represents its spectral structure. We then exploit the spatial continuity and the different spectral continuity structures of harmonic and percussive components as prior information to derive maximum a posteriori (MAP) estimates of the parameters using the expectation-maximization (EM) algorithm. Experimental results over professional musical mixtures show the effectiveness of the proposed approach.

Mots clés: Harmonic and percussive source separation; Local Gaussian model; Continuity prior.; Continuity prior

Domaines: Machine Learning [stat.ML]; Informatique

Ngoc Duong, Hideyuki Tachibana, Emmanuel Vincent, Nobutaka Ono, Rémi Gribonval, et al.. Multichannel harmonic and percussive component separation by joint modeling of spatial and spectral continuity. Acoustics, Speech and Signal Processing, IEEE Conference on (ICASSP'11), May 2011, Prague, Czech Republic. ⟨10.1109/ICASSP.2011.5946376⟩. ⟨inria-00557145v2⟩