Exploring the Extent and Impact of Playout Adjustments within VoIP Applications on the MOS Scale

In coping with best-effort service, many VoIP applications employ adaptive playout strategies. Objective methods of speech quality assessment such as the ITU-T Recommendation P.862 (also known as Perceptual Evaluation of Speech Quality PESQ) typically do not capture distortion due to playout adjustments as they match up short segments prior to analysis. Similarly, the ITU-T E-Model does not capture the effect of delay variation and uses an average delay figure in its calculations. In this paper we explore in some detail, the extent of playout adjustments within VoIP applications and assess the likely impact on Mean Opinion Score MOS. We review the impact of various factors such as Voice Activity Detection (VAD) settings and hangover thresholds on talkspurt/silence period distribution. In this context we examine the distribution of playout adjustments resulting from various playout algorithms and assess the likely impact on MOS. We show that our hybrid playout strategy which utilises synchronised time to implement an informed fixed delay playout strategy wherever possible will significantly reduce playout adjustments and any consequent MOS degradation.