A model for online assessment of speech quality over IP networks based on simplified E-Model

Speech quality detection is a research focus in the field of real-time voice communication. Aiming at the problems of the complexity of parameters required for the calculation and the non-open-source nature of algorithms in the current speech quality detection models, this paper proposes a speech quality online detection model based on the classical E-Model. The speech transmission impairments in IP networks are modeled by constructing a codec impairment model, a packet loss impairment model, and a delay impairment model, and the speech quality prediction formulas for speech codecs are obtained using subjective speech quality assessment and regression analysis methods. In this paper, the speech quality formulas for predicting the Opus and iLBC speech codecs are established using the above models. The experimental results show that the method described in this paper effectively fits the subjective listening model, extends the application range of the E-Model to the full frequency band, and supports the online assessment of the speech quality of IP speech codecs with variable bit rates, which provides strong support for improving the resistance to weak networks of IP speech service systems.