Monitoring Speech Quality in Telephone Networks: An Overview

Monitoring speech quality in telephone networks is a two-step proce-dure. First, all data which are relevant for speech quality have to be measured during network operation. Then, models can be defined which calculate quality indices on the basis of the measurements. This talk will give an overview of monitoring approaches related to speech communication quality, which is one component of the quality of a service (QoS). Intrusively or non-intrusively measured data forms an input to different types of quality prediction models. Such models are based on the perceptively based comparison between input and output signals of the transmission line (signal-based mod-els, e.g. PESQ, TOSQA, etc.), or on empirically determined relations between network planning parameters and user quality judgements (network planning models, e.g. the E-model or SUBMOD). They produce an estimate for certain quality dimensions which depend on the type of input parameters taken into account by the measurements. Various combinations between input parameters and prediction models are discussed, and limitations of the available methods are pointed out. 1