Analytical method for objective scoring of HTTP ADAPTIVE STREAMING (HAS)

We demonstrate in this paper the relationship between viewer Quality of Experience (QoE) or Mean Opinion Score (MOS) and the HTTP Adaptive Streaming Video Quality (VQ) rate. We show that inelasticity exists between the mean VQ rate and MOS scores, where multi megabit changes in VQ rates result in only modest changes in MOS scores. This is a significant network architecture design issue and potential cost benefit for operators with constrained bandwidth and access speed. We demonstrate that MOS scores are impacted by the variance (or standard deviation-(SD)) of the VQ rate more so than the mean VQ level. In fact, for Microsoft HAS changes in SD impacted the MOS by more than 2:1 over changes in mean VQ encoded rates. In other words users care much more about maintaining a consistent VQ level than they do in a higher bit rate. Finally we demonstrate that we can accurately predict MOS scores incorporating both the mean VQ level and the SD of the VQ level. Multiple analytical methods are given for predicting MOS, and are shown to have a high correlation and low error.