Speech intelligibility enhancement

Contact: Richard Hendriks

The use of speech processing devices in our society has become widespread. Examples are public address systems, mobile telephony, hearing aids and internet telephony.

In mobile situations, the user environment is typically noisy with increased intelligibility problems as a result.

To increase the speech quality and intelligibility, it is common practice to apply noise reduction and speech enhancement algorithms to the signals recorded at the far end and played back at the near end, respectively.

To reduce far-end noise, it is common to apply single or multi-microphone noise reduction algorithms at the far-end. Single-microphone methods are mostly effective for increasing the speech quality, while multi-microphone methods are able to improve the speech intelligibility as well.

At the near-end, the pre-processing is performed on the speech signal that is received from the far-end, to maintain its intelligibility when played out in the noisy near-end environment.

The repository contains Matlab code of several algorithms which we have proposed and published over the past decade.

Items under this topic

Distributed Rate-Constrained LCMV Beamforming

Software, Oct 2019

Joint near-end and far-end intelligibility enhancement

Software, Jan 2017

Optimal energy redistribution for speech enhancement

Software, Feb 2015

STOI-Optimal N-of-M Channel Selection for Cochlear Implants

Software, Aug 2012

Near-End Speech Enhancement

Pre-processing algorithm to improve speech intelligibility in noise for the near-end listener.

Software, May 2012