S. Braun, M. Torcoli, D. Marquardt, E. A. P. Habets and S. Doclo
Published in the Proc. of the International Workshop on Acoustic Signal Enhancement (IWAENC), 2014.
This paper received the best student paper award at IWAENC 2014.
Dereverberation for multichannel hearing aids is still a field of extensive research. True binaural multichannel spatial filtering techniques such as the MVDR beamformer or the multichannel Wiener filter (MWF) have the common problem, that they distort the binaural cues of the residual undesired signal components if a single desired source is assumed. A recently proposed method uses an MWF with an additional constraint to preserve the interaural coherence of a stationary diffuse noise field between the hearing aid signals. By modeling the reverberation as a time-varying diffuse field, we can achieve a reasonable amount of dereverberation and noise reduction. Therefore, the resulting filter has to be time-varying. We propose a method to preserve the interaural coherence of the undesired signal components, which are in our case reverberation and noise. It is shown, that we can preserve the coherence at the output while the reduction of reverberation and noise is still sufficiently high.
The signals are obtained from measured impulse response data from a dummy head wearing hearing aids. The source is located in front of the listener. Details of the scenario and setup are described in the submitted paper.
Listen to the first 3 signals to compare the MWF and MWF-IC. After that, you can listen to the interference component (diffuse sound + noise) only. You will hear that the MWF destroys the spatial impression of the interference component, while the MWF-IC preserves it: Listen to the envelopment of the diffuse sound.