Measuring the perceptual quality of Skype sources

The success of Skype has inspired a generation of peer-to-peer-based solutions for real-time multimedia services over the Internet. However, there lacks still a robust metric quantifying the perceptual quality of a Skype call. The widely-used PESQ (Perceptual Evaluation of Speech Quality) falls short of modeling super-wideband calls, which are characteristics of SILK Skype's codec made public in 2011. Towards a robust QoE (Quality of Experience) metric for VoIP call analysis, we propose a model, referred to as WF-Regression model, to capture the call rate and perceptual quality relationship. The model is shown through a user study that it is robust, R-square = 0.9990 and outperform PESQ modeling the quality of Skype calls, error ratio = 3.68% vs. 14.59%.