Search results

This paper aims to argue on the efficiency of Quality of Service (QoS)-based adaptive streaming with regards to perceived quality Quality of Experience (QoE). Although QoS parameters are extensively used even by high-end adaptive streaming algorithms, achieved QoE fails to justify their use in real-time streaming videos with high motion. While subjective measurements of video quality are difficult to be applied at runtime, objective QoE assessment can be easier to automate. For end-to-end QoS optimization of live streaming of high-motion video, objective QoE is a more applicable approach. This paper contributes to the understanding of how specific QoS parameters affect objective QoE measurements on real-time high-motion video streaming.

The paper approached the question through real-life and extensive experimentation using the Skype adaptive mechanisms. Two Skype terminals were connected through a QoS impairment box. A reference video was used as input to one Skype terminal and streamed on one direction. The impairment box was stressing the stream with different conditions. Received video was stored and compared against the reference video.

After the experimental analysis, the paper concludes that adaptive mechanisms based on QoS-related heuristics fail to follow unexpected changes to stream requirements. High-motion videos are an example of this variability, which makes the perceived quality sensitive to jitter more than to packet loss. More specifically, Skype seems to use if-else heuristics to decide its behavior to QoS changes. The weaknesses to high-motion videos seem to lie on this rigidity.

Due to the testbed developed, the results may be different if experiments are run over networks with simultaneous streams and a variety of other traffic patterns. Finally, other streaming clients and algorithms would contribute to a more reliable generalization.

The paper motivates video streaming engineers to emphasize their efforts toward QoE and end-to-end optimization.

The paper identifies the need of a generic adaptive streaming algorithm able to accommodate a big range of video characteristics. The effect of QoS variability to high-motion video streaming helps in modeling and design.