Publica
Hier finden Sie wissenschaftliche Publikationen aus den FraunhoferInstituten. Harmonicpercussive source separation with deep neural networks and phase recovery
 Saruwatari, H. ; Institute of Electrical and Electronics Engineers IEEE: 16th International Workshop on Acoustic Signal Enhancement, IWAENC 2018. Proceedings : 17th20th September 2018, Tokyo, Japan Piscataway, NJ: IEEE, 2018 ISBN: 9781538681510 ISBN: 9781538681503 ISBN: 9781538681527 pp.421425 
 International Workshop on Acoustic Signal Enhancement (IWAENC) <16, 2018, Tokyo> 

 English 
 Conference Paper 
 Fraunhofer IDMT () 
Abstract
Harmonic/percussive source separation (HPSS) consists in separating the pitched instruments from the percussive parts in a music mixture. In this paper, we propose to apply the recently introduced MaskerDenoiser with twin networks (MaD TwinNet) system to this task. MaD TwinNet is a deep learning architecture that has reached stateoftheart results in monaural singing voice separation. Herein, we propose to apply it to HPSS by using it to estimate the magnitude spectrogram of the percussive source. Then, we retrieve the complexvalued shorttime Fourier transform of the sources by means of a phase recovery algorithm, which minimizes the reconstruction error and enforces the phase of the harmonic part to follow a sinusoidal phase model. Experiments conducted on realistic music mixtures show that this novel separation system outperforms the previous stateofthe art kernel additive model approach.