Combining blockwise and multi-coefficient stepwise approches in a general framework for online audio source separation

Laurent S. R. Simon (1) , Emmanuel Vincent (2)

1 LIMSI - Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur
2 MULTISPEECH - Speech Modeling for Facilitating Oral-Based Communication

This article considers the problem of online audio source separation. Various algorithms can be found in the literature, featuring either blockwise or stepwise approaches, and using either the spectral or spatial characteristics of the sound sources of a mixture. We offer an algorithm that can combine both stepwise and blockwise approaches, and that can use spectral and spatial information. We propose a method for pre-processing the data of each block and offer a way to deduce an Equivalent Rectangular Bandwith time-frequency representation out of a Short-Time Fourier Transform. The efficiency of our algorithm is then tested for various parameters and the effect of each of those parameters on the quality of separation and on the computation time is then discussed.

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

Laurent S. R. Simon, Emmanuel Vincent. Combining blockwise and multi-coefficient stepwise approches in a general framework for online audio source separation. [Research Report] RR-8766, Inria. 2015, pp.18. ⟨hal-01186948⟩