Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

A geometrically constrained independent vector analysis algorithm for online source extraction

: Khan, Affan-Hassan; Taseska, Maja; Habets, Emanuël A.P.


Vincent, E.:
Latent variable analysis and signal separation. 12th international conference, LVA/ICA 2015 : Liberec, Czech Republic, August 25 - 28, 2015; Proceedings
Cham: Springer International Publishing, 2015 (Lecture Notes in Computer Science 9237)
ISBN: 978-3-319-22481-7 (Print)
ISBN: 978-3-319-22482-4 (Online)
International Conference on Latent Variable Analysis and Signal Separation (LVA/ICA) <12, 2015, Liberec/Czech Republic>
Fraunhofer IIS ()

In this paper, an online constrained independent vector analysis (IVA) algorithm that extracts the desired speech signal given the direction of arrival (DOA) of the desired source and the array geometry is proposed. The far-field array steering vector calculated using the DOA of the desired source is used to add a penalty term to the standard cost function of IVA. The penalty term ensures that the speech signal originating from the given DOA is extracted with small distortion. In contrast to unconstrained IVA, the proposed algorithm can be used to extract the desired speech signal online when the number of interferers is unknown or time varying. The applicability of the algorithm in various scenarios is demonstrated using simulations.