Speech quality and the E-model

The E-model is a quality prediction model for network planning purposes. Its use is going to be recommended by the ITU-T. In comparison with both listening-only and conversation tests the model often shows high predictive power. Accordingly, network planners have high confidence in its applicability. However, until now it is not yet discretely understood which aspects of quality are actually covered by E-model predictions. To name only two, uncertainties do exist, e.g., for the mode of communication (listening only or conversation) and user-related factors. In the talk, a schematic is presented which covers, besides speech-communication-related elements (one-way voice transmission quality, communication effectiveness, ease of communication) also human-factor-related and service-related elements (comfort, costs). It is shown that matters are simplified by exclusively using the MOS as the basis for a thorough verification of the model. In the course of verification, measures of user acceptance, expectation, etc., are needed in order to understand how human- and service-related features influence the users’ perception of the quality of service. An approach is presented to obtain subjective judgments other than MOS on a scale which is closely related to the scale of the model, namely the transmission rating factor R.