Parametric and signal-based models for wideband speech quality


Motivation & Project Description

• Development of both parametric and signal-based models for the prediction of wideband speech quality:
• Motivation: In order to reduce costs for subjective tests, models are needed for the quantification of wideband speech quality in an end-to-end scenario, both for network planning as well as for network monitoring.
  • Extension of the E-Model towards wideband speech, extending the approach known from narrowband network planning.
  • Extension of the E-Model towards wideband speech, based on perceptual dimensions.
  • Development of a signal-based model for estimating overall quality as well as perceptually relevant dimensions.

Outcome:

• Models and algorithms for estimating narrowband and wideband speech quality
• ITU-T contributions and recommendations (ITU-T SG 12, Q.8/12 and Q.9/12)
• Scientific publications

Time Frame: 05/2007-12/2008

T-labs Team Members: Nicolas Côté, Sebastian Möller, Alexander Raake, Marcel Wältermann

Students: Ulrike Stiefelhagen

Partners: France Télécom R&D

Funding by: Deutsche Telekom Laboratories, Deutsche Forschungsgemeinschaft (DFG), France Télécom R&D