Evaluation de la qualité vocale dans les télécommunications
Marie Guéguin, Vincent Barriac, Valérie Gautier-Turbin, Régine Le Bouquin-Jeannès, Gérard Faucon.
This paper is a review of the methods for speech quality assessment. Subjective methods involve human subjects testing systems in various network conditions and voting on an opinion scale. The scores obtained for each condition are averaged to get a mean opinion score (MOS). These subjective tests are the only way to assess perceived speech quality, but they are complex, cost- and time-consuming. Consequently objective methods have been introduced to predict the speech quality as perceived by users. Here, objective methods are classified depending on the context they deal with. This review of objective methods shows a lack of model in the conversational context. Then we propose an objective model of the conversational speech quality, built on a combination of objective models of the listening and talking speech qualities and the delay.