Dereverberation of speech signals

Researchers: Felicia Lim and Patrick A. Naylor

Speech signals recorded in an enclosed environment are subject to degradation due to reverberation. This is the effect of the speech signal reflecting off surfaces in the room, causing multiple delayed and attenuated copies of itself to be superimposed in the recorded signal.

This has the unwanted consequence of decreasing the quality of speech perception and reducing the performance of other speech signal processing algorithms such as automatic speech recognition.

The degradation is more pronounced when the recording is done in hands-free mode. In this case, the speaker is further away from the microphone and hence, there is a smaller difference between the intensities of signals from the direct and reflected paths.

Dereverberation aims to cancel or suppress the reverberation effect and is of key importance to applications requiring hands-free modality. Some examples include hearing aids, telecommunication devices and voice-controlled systems.

This project investigates one method of dereverberation, which aims to perform multichannel equalization of an estimated acoustic impulse response. In particular, techniques within the framework of channel-shortening based algorithms are investigated and evaluated.

Relevant publications:

  • F. Lim, P. A. Naylor: Statistical Modelling of Multichannel Blind System Identification Errors. In: Proc. Intl. Workshop Acoust. Signal Enhancement (IWAENC), Juan les Pins, France, 2014.
  • M. R. P. Thomas, F. Lim, I. J. Tashev, P. A. Naylor: Optimal Beamforming as a Time Domain Equalization Problem with Applications to Room Acoustics. In: Proc. Intl. Workshop Acoust. Signal Enhancement (IWAENC), Juan les Pins, France, 2014.
  • F. Lim, W. Zhang, E. A. P. Habets, P. A. Naylor: Robust Multichannel Dereverberation using Relaxed Multichannel Least Squares. In: IEEE/ACM Trans. Audio, Speech, Lang. Process., 22 (9), pp. 1379-1390, 2014.
  • F. Lim, M. R. P. Thomas, P. A. Naylor: MINTFormer: A spatially aware channel equalizer. In: Proc. IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), New Paltz, New York, USA, 2013.
  • F. Lim, P. A. Naylor: Robust speech dereverberation using subband multichannel least squares with variable relaxation. In: Proc. European Signal Processing Conference (EUSIPCO), Marrakech, Morocco, 2013.
  • F. Lim, P. A. Naylor: Robust low-complexity multichannel equalization for dereverberation. In: Proc. IEEE Intl. Conf. on Acoustics, Speech and Signal Processing (ICASSP), Vancouver, Canada, 2013.
  • F. Lim, P. A. Naylor: Relaxed multichannel least squares with constrained initial taps for multichannel dereverberation. In: Proc. Intl. Workshop Acoust. Signal Enhancement (IWAENC), Aachen, Germany, 2012.

Audio examples

Contact us

Address

Speech and Audio Processing Lab
CSP Group, EEE Department
Imperial College London

Exhibition Road, London, SW7 2AZ, United Kingdom

Email

p.naylor@imperial.ac.uk