This paper proposes a model for the computational assessment of IP-based audio quality in the absence of a reference signal. It is similar in type to the E-model for speech signals and includes two types of impairments causing quality degradation: audio stream encoding and network packet loss. Effects of the impairments were parameterized using a subjective experiment. Because subjective tests for impairment quantification are complex and time consuming, the paper also proposes a methodology for using the objective full-reference PEAQ algorithm to quantify parameters for the developed non-reference model.
展开▼
