E-model supported switching between narrowband and wideband speech quality

Forthcoming next generation networks (NGNs) introduce a new trend of integration into one all-IP platform. This process will decouple VoIP services from the network providers, introducing new service distribution aspects. VoIP devices will be enabled to seamlessly switch between wideband and narrowband speech operators and roam across diverse networks on demand, even during active calls. To design user-friendly speech quality switching posed by the NGNs, a reliable non-intrusive quality prediction model for quality monitoring will be required. Until now, none of the existing quality prediction models has been fully validated for use in NGNs. This prohibits quality prediction, when switching between wideband and narrowband speech codecs is necessary. In this paper, we validate the applicability of the e-Model for NGNs, targeting the situations when the speech bandwidth must be changed. We propose an additive codec switching impairment factor I
_esw
that improves the e-Model estimation for the new quality phenomenon. In this way, the achieved e-model correlation r = 0.93 enables VoIP quality monitoring of dynamic VoIP quality profiles, which is required to provide maximal user satisfaction in NGNs.