In this article, we assessed the influence of quality fluctuations during
HTTP Adaptive Streaming to mobile devices. In this work, we
went beyond existing state-of-the-art subjective studies and recommended
assessment methodologies by conducting novel experiments
in more ecologically valid environments using long duration video
content. As such, we extended the scope of existing subjective studies.
Our experiments were conducted on iPad devices in subjects’ own
home environment. Furthermore, we also included different video
content types.
Based on different simulated HAS scenarios, we assessed the influence
of bitrate fluctuations and video stallings on quality perception.
In this respect, we also considered the influence of switching frequency
and switching range. Using continuous quality evaluation we
were able to track end-users’ behaviour in case of changing video
quality.
We found that test subjects are able to track the simulated quality
fluctuations. Corresponding to earlier research findings, our results
show that, in general, video stallings should be avoided at all times
and that gradual quality changes have a smaller influence on quality
perception compared to instantaneous switches.
In this study, we found significant influences of video content and
the range of quality switches on the average rating behaviour. In
our case, quality level switches are only perceived in high motion
sequences or in case quality level switches occur between highest
and lowest quality of the video. Limited noticeable differences
were observed between high and mid-quality switches during video
playback.
The results of this study can be used to implement more intelligent
adaptation strategies in the HAS clients in order to ensure and maintain
QoE towards the end-users at all times. In this respect, as part
of the adaptation strategies, both video stallings and high multi-range
quality switches should be avoided.
Concerning conducting our experiments in more ecologically valid
testing environments, we found that our results deviate less from
experiments in controlled environments compared to some of our earlier
research on the influence of assessment environment on quality
perception. However, as we relied on continuous quality evaluation,
test subjects remain concentrated on actively evaluating video quality
throughout the entire experiment which hampers an immersive
experience.