Video streaming services account for the majority of today’s traffic on the Internet. Although the data transmission rate has been increasing significantly, the growing number and variety of media and higher quality expectations of users have led networked media applications to fully or even over-utilize the available throughput. HTTP Adaptive Streaming (HAS) has become a predominant technique for multimedia delivery over the Internet today. However, there are critical challenges for multimedia systems, especially the tradeoff between the increasing content (complexity) and various requirements regarding time (latency) and quality (QoE). This thesis will cover the main aspects within the end user’s environment, including video consumption and interactivity, collectively referred to as player environment, which is probably the most crucial component in today’s multimedia applications and services. We will investigate the methods that can enable the specification of various policies reflecting the user’s needs in given use cases. Besides, we will also work on schemes that allow efficient support for server-assisted, and network-assisted HAS systems. Finally, those approaches will be considered to combine into policies that fit the requirements of all use cases (e.g., live streaming, video on demand, etc.).
4. HTTP Adaptive Streaming (HAS)
4
Video
segmentation
Video
encoding
...
...
...
Server Client/ HAS player
Adaptive bitrate
algorithm
Throughput
estimation
Playout buffer
Video decoding
...
HTTP GET requests
Version 3
Version 2
Version 1
Internet
Throughput
Time
7. Client side
7
Client
Adaptive bitrate
algorithm
Bentaleb, et al. "A survey on bitrate adaptation schemes for streaming media over
HTTP." IEEE Communications Surveys & Tutorials 21.1 (2018): 562-585.
Markov Decision
Process based
S2
S0
S1
Artificial Intelligence
based
AI
Hybrid
Throughput based
Buffer based
8. Emerging and future networking approaches
8
Internet
...
HTTP/2
HTTP/3
SDN
5G
TCP
TCP
Multipath
TCP
10. Research questions
10
● RQ1: How to provide a generic mechanism for HAS players that meets
customer needs?
○ Tradeoff: content, quality, and time for each use case (e.g., live streaming,
and video on demand).
● RQ2: How to enable efficient support for server-/network assisted HAS?
○ Sharing server-/network information to/among HAS players to ensure
quality fairness among multiple clients.
11. Research questions
11
● RQ3: How to add support for emerging/future networking aspects and paradigms?
○ Utilizing their features to improve HAS player’s performance.
● RQ4: How to enable client-based quality enhancement options for HAS players?
○ Applying Deep Neural Networks such as super-resolution ones to avoid the dependence
on the network condition.
● RQ5: How to integrate advanced analytics options and various predictions models
for HAS players?
○ Utilizing machine learning-based methods such as Long Short Term Memory.
13. Methodology
13
Analyze user needs.
Research state-of-
the-art.
State requirements
Describe behaviors
and technical
characteristics.
State specifications
Design and implement
possible solutions.
Design and implement
Conduct extensive
experiments and
make comparison.
Test
Follow design paradigm introduced by Association
for Computing Machinery (ACM) [1].
[1] D. E. Comer, David Gries, Michael C. Mulder, Allen Tucker, A. Joe Turner, and Paul R. Young. Computing as a discipline. Communications of the
ACM, 32(1):9– 23, January 1989
15. H2BR: An HTTP/2-based Retransmission
Technique to Improve the QoE of Adaptive Video
Streaming
15
● Research goals:
○ Improving low-quality segments
○ Filling the quality gaps
Nguyen, M., Timmerer, C. and Hellwagner, H., 2020, June. H2BR: An HTTP/2-based retransmission technique to improve
the QoE of adaptive video streaming. In Proceedings of the 25th ACM Workshop on Packet Video, pages 1-7, 2020.
16. H2BR: An HTTP/2-based Retransmission
Technique to Improve the QoE of Adaptive Video
Streaming
16
● Our approach
○ Retransmitting downloaded segments with
higher quality versions.
○ Utilizing HTTP/2 features (multiplexing,
server push, stream priority, termination).
● Experimental results
○ - 70% lowest-quality segments
○ + 13% QoE
Nguyen, M., Timmerer, C. and Hellwagner, H., 2020, June. H2BR: An HTTP/2-based retransmission technique to improve
the QoE of adaptive video streaming. In Proceedings of the 25th ACM Workshop on Packet Video, pages 1-7, 2020.
17. WISH: User-centric Bitrate Adaptation for HTTP
Adaptive Streaming on Mobile Devices
17
● Observation
Nguyen, M., Cetinkaya, E., Hellwagner, H. and Timmerer, C., WISH: User-centric Bitrate Adaptation for HTTP Adaptive
Streaming on Mobile Devices. In 2021 IEEE 23rd Int’l. Workshop on Multimedia Signal Processing (MMSP). IEEE, 2021
More transferred data
(high data cost)
More download time
(high buffer cost)
Higher quality
(less quality cost)
Download
high-bitrate
segment
18. WISH: User-centric Bitrate Adaptation for HTTP
Adaptive Streaming on Mobile Devices
18
● Idea
○ The overall cost of each representation is the weighted sum of
Throughput cost, Buffer cost, and Quality cost
○ Lowest overall cost representation is selected.
● Results
○ - 36% data usage
○ + 18% QoE
Nguyen, M., Cetinkaya, E., Hellwagner, H. and Timmerer, C., WISH: User-centric Bitrate Adaptation for HTTP Adaptive
Streaming on Mobile Devices. In 2021 IEEE 23rd Int’l. Workshop on Multimedia Signal Processing (MMSP). IEEE, 2021
19. Ongoing and Future Work
19
● RQ 2
○ Define a notion of QoE fairness
○ Design optimization models with the input as server-/network information
to maximize QoE fairness
● RQ 4
○ Apply Deep Neural Networks such as super-resolution to improve the QoE.
● RQ 5
○ Work on QoE prediction models
○ Investigate appropriate client metrics to improve QoE