QoE-aware priority marking and traffic management for H.264/SVC-based mobile video delivery
2013
Mobile network operators experience a rapid increase of traffic in their networks that is mainly driven by the growing popularity of mobile video applications. They partly compensate this traffic increase by improved transmission technology and network densification. Supporting videos at high quality during peak traffic, however, does not scale economically. Instead, means are required for the operator to adapt the bitrate of the videos to the current network situation. Scalable Video Coding (SVC) provides a convenient way to adapt the bitrates of video streams in the network by dropping bit stream layers. From resource utilization point of view, the limitation of SVC is that the layer structure does not allow an operator to compare the importance of two layers of different videos with respect to their contribution to the user-perceived video quality. In existing solutions, the limitation is resolved by implementing sophisticated optimization algorithms at network bottlenecks for optimal rate adaptation among multiple videos. We take a different approach to eliminate the overhead of signaling and implementations introduced by the optimizations at bottlenecks while targeting at the same optimization goal. We propose a QoE-aware priority marking algorithm for SVC-based video streaming. In our approach, SVC layers are mapped to a finite number of priorities which are marked in video packets to be transmitted with video streams. This mapping takes into account the data rate and the quality contribution of a layer among layers of a group of videos. At network bottlenecks, rate adaptation simply follows the pre-determined priorities without further optimization needed. Our results show that the QoE-aware priority marking algorithm indeed overcomes the limitations of existing SVC packet marking and has the potential to achieve a QoE-optimal resource utilization in wireless access networks.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
10
References
13
Citations
NaN
KQI