E-model modification for case of cascade codecs arrangement

Abstract

Speech quality assessment is one of the key matters of voice services and every provider should ensure adequate connection quality to end users. Speech quality has to be measured by a trusted method and results have to correlate with intelligibility and clarity of the speech, as perceived by the listener. It can be achieved by subjective methods but in real life we must rely on objective measurements based on reliable models. One of them is E-model that we can consider as mainly adopted method in IP telephony. This method is based on evaluation of transmission path impairments influencing speech signal, especially delays and packet losses. These parameters which are common in IP network can affect dramatically speech quality. In this article, a new modification of E-model, that takes into consideration the cascade codecs arrangement, is presented. The proposed a correction function improves the current computational non-intrusive approach that is described in recommendation ITU-T G.107, so-called E-model.