scope:

The present document provides information about the impact of specific types of packet loss and coding on speech quality predictions provided by ITU-T Recommendation P.862 [i.6] and P.563 [i.9] models, when both naturally-produced and synthesized speech are used. The variability of predictions with respect to the type of signal used (naturally-produced or synthesized) and loss conditions as well as their accuracy were assessed by comparing the predictions with subjective assessments.

The results indicate some implications for designers of speech communication systems.

It has to be emphasized that none of the instrumental algorithms investigated here (P.862 and P.563) were validated for synthesized speech. The presented analysis is a use case which is out-of-scope for these algorithms.