Predicting the perceptual service quality using a trace of VoIP packets

Abstract

We present an instrumental approach on how to assess the perceptual quality of voice transmissions in IP-based communication networks. Our approach is end-toend and uses combinations of common codecs, loss concealment algorithms, playout schedulers, and ITU's quality assessment algorithms Å-Model and PESQ. It is the first method that takes the impact of playout rescheduling and non-random packet loss distributions into account. Non-random packet losses occur if a ratedistortion optimized multimedia streaming algorithm forwards packets dependent on the packets' importance. Our approach is implemented in open-source software. We have conducted formal listening-only tests to verify the accuracy of our quality model. In the majority of cases, the human test results show a high correlation with the calculated predictions. © Springer-Verlag Berlin Heidelberg 2004.