DREAMS – Dereverberation and Reverberation of Audio, Music and Speech

Room reverberation, as is familiar in particular in large buildings such as churches, railway stations or concert halls, considerably influences the speech intelligibility and quality of speech and music signals. Whereas reverberation can increase the perceived quality of music signals, the intelligibility of speech signals is often significantly impaired. This is true both for perception by human hearing as well as for technical systems, e.g. machine speech recognition. In the Marie-Curie Initial Training Network »DREAMS (Dereverberation and Reverberation of Audio, Music, and Speech)«, twelve European partners from the fields of industry and research are investigating the influence of room reverberation on music and speech signals. In addition to carrying out modeling and evaluation, the partners are also developing signal processing strategies designed to remove reverberation components.

Responsibilities of Fraunhofer IDMT

© Fraunhofer IDMT/Daniel Schmidt
Demonstrator speech enhancement

The Division Hearing, Speech and Audio Technology of Fraunhofer IDMT is responsible in the project for carrying out research on algorithms for dereverberation and noise reduction in order to improve the quality of speech signals. To assess the quality of reverberant and dereverberated signals, objective quality measures based on characteristics and models of the human auditory system are developed and applied.

Partners

  • Katholieke Universiteit Leuven, Leuven, Belgium
  • Aalborg Universitet, Aalborg, Denmark
  • Imperial College of Science Technology and Medicine, London, UK
  • Carl von Ossietsky Universität, Oldenburg, Germany
  • NXP Semiconductors, Leuven , Belgium
  • Oticon A/S, Smorum, Denmark
  • Bang and Olufsen, Struer, Denmark
  • Spinvox, Marlow, UK
  • Nuance Communications, Aachen, Germany
  • Televic Group NV, Izegem, Belgium
  • CEDAR Audio Ltd, Cambridge, UK

Duration

2013-2016

Project website