Adaptive Time-Domain Blind Separation of Speech Signals

0348108 - ÚTIA 2011 RIV DE eng C - Konferenční příspěvek (zahraniční konf.)
Málek, J. - Koldovský, Zbyněk - Tichavský, Petr
Adaptive Time-Domain Blind Separation of Speech Signals.
Lecture Notes in Computer Science. Vol. 6365. Heidelberg: Springer-Verlag, 2010 - (Gavrilova, M.; Kumar, V.; Mun, Y.; Tan, C.; Gervasi, O.), s. 9-16. ISBN 978-3-642-15994-7.
[Latent Variable Analysis and Signal Separation. St. Malo (FR), 27.09.2010-30.09.2010]
Grant CEP: GA MŠMT 1M0572; GA ČR GA102/09/1278
Grant ostatní: GA ČR(CZ) GA102/08/0707
Výzkumný záměr: CEZ:AV0Z10750506
Klíčová slova: blind source separation * speech * convolutive mixture * adaptive algorithms
Kód oboru RIV: BB - Aplikovaná statistika, operační výzkum
http://library.utia.cas.cz/separaty/2010/SI/tichavsky-adaptive time-domain blind separation of speech signals.pdf

We present an adaptive algorithm for blind audio source separation (BASS) of moving sources via Independent Component Analysis (ICA) in time-domain. The method is shown to achieve good separation quality even with a short demixing filter length (L = 30). Our experiments show that the proposed adaptive algorithm can outperform the off-line version of the method (in terms of the average output SIR), even in the case in which the sources do not move, because it is capable of better adaptation to the nonstationarity of the speech.
Trvalý link: http://hdl.handle.net/11104/0188720