Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Coherence-aware stereophonic residual echo estimation

: Valero, Maria Luis; Yildiz, Ilkay; Mabande, Edwin; Habets, Emanuël A.P.


Bellegarda, J. ; IEEE Signal Processing Society:
Hands-Free Speech Communications and Microphone Arrays, HSCMA 2017. Proceedings : March 1-3, 2017, San Francisco, California, U.S.A.
Piscataway, NJ: IEEE, 2017
ISBN: 978-1-5090-5925-6
ISBN: 978-1-5090-5924-9
ISBN: 978-1-5090-5926-3
Workshop on Hands-Free Speech Communication and Microphone Arrays (HSCMA) <5, 2017, San Francisco/Calif.>
Fraunhofer IIS ()

The performance of stereophonic acoustic echo cancelers is highly dependent on the coherence between the loudspeaker signals. If this is high, convergence is slow, during which audible residual echoes remain. Residual echoes can be reduced by applying a residual echo suppressor to the canceler’s output. Yet, in order to obtain the post- filter gains, it is necessary to estimate the residual echoes first, and the performance of stereophonic residual echo estimators also de- pends on the loudspeaker signals’ coherence. In this paper, a stereo- phonic residual echo estimator is proposed that takes the coherence between the signals into account. By doing this, it is possible to reliably estimate the residual echoes, even if the loudspeaker signals are completely coherent.