Objective End-to-End ('Mouth to Ear') Conversational Speech Quality Tests for VoIP Scenarios

摘要

From the speech quality point of view the differentiation between terminals and network in communications over IP is no longer possible. Consequently the overall speech quality assessment has to take this into account and requires end-to-end tests. Suitable test setups including the terminals acoustics using artificial head technology as a close to reality interface are introduced. In a second part the influence of various subjectively relevant parameters on speech quality is discussed. Correlated objective parameters like delay, echo, double talk capability, listening speech quality and parameters determining background noise transmission quality are described. Appropriate analysis methods are given. The discussion points out the influence of delay on conversation dynamics impairments and its influence on echo perception, because the expected delay in VoIP scenarios is probably higher than typically recommended for telephone conversations. Optimization criteria are introduced for implemented echo cancellers as well as test methods to assess the one-way speech sound quality, double talk performance and background noise transmission.