2013年3月14日学术报告（电工系）-Internet视频服务的用户体验建模理论

联系人：电工系李翔

 

 

报告摘要：The video streaming services take up more than half of today's Internet traffic, and the traffic volume continues to increase very fast. It becomes a very challenging task for content providers and Internet service providers (ISPs) to guarantee satisfactory user perception, (also called Quality of Experience (QoE)). The QoE of streaming service is often degraded by frequent playback interruptions. To mitigate the interruptions, the media player prefetches some streaming packets before starting playback, at a cost of delay. An interesting category of problems are how many packets should be prefetched beforehand so as to balance the tradeoff between the smoothness of video playback and the waiting time. We build a theory of QoE for video streams transferred over wireless and wired networks. Our purpose is to provide mathematical models to compute the waiting time, and the probability and frequency of playback interruptions. In this series of works, we will show our theory can be used to predict QoE of streaming services ranging i) from one flow to multiple dynamic flows; ii) from Poission arrival to Markov modulated arrival processes and iii) from static video quality to dynamic video quality. Our theory provides deep understanding of the delicate QoE tradeoffs and helps the providers to configure the content prefetching in a better way.