Op­ti­mi­zing Speech In­tel­li­gi­bi­li­ty in Noi­sy En­vi­ron­ments Using a Sim­ple Mo­del of Com­mu­ni­ca­ti­on

Modern communication technology allows a user to communicate from almost
anywhere to almost anywhere. However, interfering sources in the environment of
the talker at the far-­‐end and the listener at the near-­‐end often affect the ability of
the parties to communicate. Conventional systems for noise reduction and speech
intelligibility enhancement typically treat the processing at the far-­‐end and the near-­‐
end as two independent problems.
To quantify and optimize the intelligibility of noisy speech, we recently introduced a
speech intelligibility model based on mutual information. This model takes the noise
inherent in the speech production process into account at a fixed SNR (i.e., scaling
independent).  Such a constant production SNR has a significant effect on a power
constrained communication system. That is, the usefulness of a particular
communication channel “saturates” near the production SNR. The resulting
intelligibility predictor resembles heuristically derived classical measures of
intelligibility such as the AI and the SII.
In this presentation we will review our proposed mutual information-­‐based speech
intelligibility model and demonstrate its potential. We will relate it to classical
measures of intelligibility and will show its potential for speech intelligibility
prediction. Furthermore, we will show that this model can be used to derive a multi-­‐
microphone processor that is jointly optimal with respect to interfering sources at
the far-­‐end as well as at the near-­‐end.

Ri­chard C. Hen­driks

Richard C. Hendriks obtained his M.Sc. and Ph.D. degrees (both cum laude) in electrical engineering from Delft University of Technology, Delft, The Netherlands, in 2003 and 2008, respectively. From 2003 till 2007, he was a Ph.D. Researcher at Delft University of Technology, Delft, The Netherlands. From 2007 till 2010, he was a Postdoctoral Re- searcher at Delft University of Technology. Since 2010, he has been an Assistant Professor in the Signal and Information Processing Lab of the faculty of Electrical Engineering, Mathematics and Computer Science at Delft University of Technology. In the autumn of 2005, he was a Visiting Researcher at the Institute of Communication Acoustics, Ruhr-University Bochum, Bochum, Germany. From March 2008 till March 2009, he was a Visiting Researcher at Oticon A/S, Copenhagen, Denmark. His main research interests are digital speech and audio processing, including single-channel and multi-channel acoustical noise reduction, speech enhance- ment, and intelligibility improvement.