The research in the SMS4PA project is aimed at signal processing algorithms which improve the intelligibility of spoken messages in public address systems. More specifically, this means that we want to be able to (acoustically) modify the speech signals in such a way that these retain better intelligibility when presented in a noisy environment. During the project, two different approaches will be followed to obtain this better intelligibility: ‘Dynamics Processing’ and ‘Clear Speech’.
Using dynamics processing [1], in a first step a fast-acting time-varying and signal dependent amplification will be applied to the signal in such a way that each phoneme individually is maximally amplified (so each phoneme is optimally audible). In a next phase, this will be supplemented with frequency dependent amplification, which allows an adaptation of the signal to the noisy acoustic environment.
The Clear Speech approach of this project encompasses that we will transform the speech signal in such a way that it seems that the speaker is trying (very) hard to be more intelligible. This will be obtained by non-uniform time scaling of the speech (corresponds to speaking more slowly) [2], and by modification of the formants of the signal. Very important here is that the speech modifications don’t (or minimally) affect the naturalness of the speech.

[1] Udo Zölzer, Ed., DAFx - Digital Audio Effects, chapter 5, pp. 95–105, John Wiley & Sons, 2002.
[2] W. Verhelst, D. Van Compernolle and P. Wambacq. 'A Unified View on Synchronized Overlap-Add Methods for Prosodic Modification of Speech.' In Proc. International Conference on Spoken Language Processing, volume II, pages 63-66, Beijing, China, October 2000.
Effective start/end date1/01/0331/12/04

    Research areas

  • electricity

    Flemish discipline codes

  • Mathematical sciences
  • Electrical and electronic engineering
  • Languages and literary studies
  • Basic sciences

ID: 2995762