Evaluation of Instrumental Quality Measures for Wideband-Transmitted Speech

Konferenz: Sprachkommunikation 2008 - 8. ITG-Fachtagung
08.10.2008 - 10.10.2008 in Aachen, Germany

Tagungsband: Sprachkommunikation 2008

Seiten: 4Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Côté, Nicolas; Gautier-Turbin, Valérie (France-Télécom RD, Lannion, 2 av. Pierre Marzin, 22300, France)
Côté, Nicolas; Möller, Sebastian (Deutsche Telekom Laboratories, Berlin Institute of Technology, Berlin, Ernst-Reuter Platz 7, Germany)

are evaluated in this paper. Many conditions are assessed including narrow-band (NB) and wideband (WB) speech codecs, transmission errors and noisy conditions. The predicted quality ratings were obtained from three kinds of speech quality measures: (i) time-domain, (ii) frequency-domain and (iii) perception-based measures. Two parameters are used to evaluate the measures: the Pearson correlation coefficient and the prediction error. This paper presents a comparison of three decades of development in the field of instrumental measures of speech quality from the signal-to-noise ratio to the model standardized in ITU-T Recommendation P.862. Results show that perceptual measures provide more reliable estimations of the auditory scores than simple timeand frequency-domain measures. However, these last two measures seem to be coherent with subjects’ ratings for some specific degradations.