QoE in DASH

Int. J. Advanced Networking and Applications
Volume: 09 Issue: 04 Pages: 3515-3522(2018) ISSN: 0975-0290
3515
QoE in DASH
Koffka Khan
Department of Computing and Information Technology The University of the West Indies, Trinidad and Tobago, W.I
Email: koffka.khan@gmail.com
Wayne Goodridge
Department of Computing and Information Technology The University of the West Indies, Trinidad and Tobago, W.I
Email: wayne.goodridge@sta.uwi.edu.com
-------------------------------------------------------------------ABSTRACT---------------------------------------------------------------
Over recent years there has been a considerable shift, from quality of service (QoS) to quality of experience (QoE),
when evaluating video delivery across networks. Hence, we first explore the need for this shift towards user-QoE
in the video delivery ecosystem. Further, we investigate major QoE metrics researchers use in the evaluation of
DASH users. We point out a huge problem with DASH beginning with its transport layer protocol. DASH utilizes
Transmission control protocol (TCP) as the transport layer protocol. Thus, we give an overview of the mechanism
of Transmission Control Protocol (TCP) and two mechanisms greatly impacting the streaming process: (1) TCP
congestion mechanism and (2) TCP Fast Start. This leads us to investigate the impact of these TCP mechanisms on
DASH players and consequently user-QoE.
Keywords - QoS; user-QoE; video; TCP; transport layer; DASH; congestion; fast start.
---------------------------------------------------------------------------------------------------------------------------------------------------
Date of Submission: Jan 26, 2018 Date of Acceptance: Feb 06, 2018
---------------------------------------------------------------------------------------------------------------------------------------------------
I. INTRODUCTION
The past ten year growth of Internet services creates a
need for an increase in multimedia delivery across a
standard medium. Consequently, today's Internet is
abundant with applications utilizing video. Old-fashioned
RTSP [27] streaming eventually evolves to HTTP-based
streaming protocols. This shift leads to a better user-
perceived QoE. Progressive download utilizes HTTP as a
protocol and succeeds traditional streaming. Today, HTTP
adaptive video streaming is the de facto standard. Real-
time multimedia delivery possesses tight latency
constraints, and data arriving too late is essentially useless.
To create the illusion of motion, video frames play
between 24-30 frames per second (fps). Each client media
compression creates interdependence between packet
contents and codecs. Thus, packet losses and late arrivals
of video data can be detrimental. A combination of the
inherent nature of network environments and transport
protocol behaviors presents multimedia delivery
challenges.
MPEG develops a HAS standard in 2012, Dynamic
Adaptive Streaming over HTTP (DASH), [31], [13], [29].
It is also known as MPEG-DASH. DASH aims to provide
video streaming service to users with dynamic network
conditions and heterogeneous devices, with no interrupts.
MPEG-DASH uses an application layer Adaptive BitRate
(ABR) algorithm. The main goal of ABR algorithms is to
prevent player’s playout buffer under-run, while
maximizing user-QoE, [34]. DASH ABR algorithms
achieve this by adapting to dynamic changes in network
conditions. The key differences between DASH and
earlier protocols for multimedia streaming are: (a) Unlike
earlier UDP-based schemes, DASH is built on top of TCP
transport, (b) The player drives the algorithm. Depending
on its ABR, the player typically requests video bitrates
based on observed network conditions, hence regulating
the server's transmission rate and (c) DASH requests and
receives video data in terms of multi-second video
segments, instead of a continuous stream of video packets.
The player algorithm dynamically adapts the quality level
in DASH-based systems. This adjustment focuses on the
current network conditions. HAS addresses the
weaknesses of RTSP, progressive download and regular
segmented streaming. However, the server pre-processes
media content. The media content is split into several
small segments of a certain length, as in segmented
streaming. The server now encodes each segment into a
number of predefined qualities. The different versions of
the segments are then distributed over one or more media
servers (HTTP web servers). The video player decodes
each representation of the segments individually. Also, the
segment length determines the shortest video duration
after which a quality adjustment can occur. Usually, this is
2 or 4 seconds in length.
This work consists of four sections. Section II discusses
the need for quality of experience (QoE) in adaptive video
streaming networks. Section III presents QoE metrics.
Section IV gives an outline of TCP congestion and fast
start mechanisms. Section V gives experimental evidence
supporting the need for QoE improvements in present day
DASH. Section VI explains the result and how the TCP
mechanisms negatively affect user-QoE. Finally, the
conclusion is given in Section VII.
II. THE NEED FOR QUALITY OF EXPERIENCE
(QOE)
Current needs for the design of interactive, personal and
socially connected platforms makes the video delivery
landscape a rapidly evolving ecosystem. The Internet is a
ubiquitous high-speed communication network supporting
such evolution. However, the Internet does not provide
any Quality of Service (QoS) or Quality of Experience
(QoE) guarantees. Thus, applications dynamically adapt
their requirements to the available QoS or QoE level.

3516
Quality of experience (QoE) quickly emerges as a
multidisciplinary field. QoE formulates itself from social
psychology, cognitive science, economics and engineering
science, cf. Figure 1. It focuses on understanding overall
human quality requirements, [12]. The design of QoE
culminates human expectations and quality needs. It’s
traditional counterpart is Quality of Service (QoS), [35],
[6], [15], [1]. QoS parameters ensure service quality to end
users. However, QoE incorporates aesthetic and
parsimonious needs of individuals. QoS lacks these scales.
Research in, [33] defines Quality of Experience (QoE):
"the overall acceptability of an application or service, as
perceived subjectively by the end user."
Fig. 1. Quality of Experience in adaptive video streaming
Measurable objective performance of technical system
entities and subjective testing with people classifies
network service evaluation scales. User-perceived QoS
scales are subjective. They gather opinions from
participants who rate the quality or change of an
experience with an entity. Advantages of this approach are
(1) it is very user-centric, (2) tests are well documented,
(3) it contains standard repeatable components and (4) it
can easily be compared to technical system performance.
However, disadvantages, such as possible defects of
individual human perceptual traits (example, ear infection
during testing) and unconscious psychological factors, [2],
may cause difficulties with this approach.
QoE gathers measures from end users, whereas QoS
measures technical patterns affecting system performance.
Subjective QoE extends beyond user-perceived media
quality. It measures user satisfaction, [9]. Subjective user
perception mainly focuses on questionnaires and rating
scales, [3]. However, user behavior relying only on
opinion is not reliable. Usually, objective measures
concern the system and not the end users. QoS tools
collect and monitor these objective measures. For
example, in a network the buffer levels can be monitored.
Here, these indirect technical measures infer QoE. Thus,
user measures validate the relationship between technical
measures and user behavior. However, user-QoE are also
affected by other factors, cf. Figure 2.
Fig. 2. Factors affecting QoE in adaptive video streaming
Computational models of media quality are objective
because they model measurable technical parameters, [22].
They access cumulative variations in technical parameters,
including those that affect quality. However, the value of
these models reduces, as they must be continually
validated against user test data, with the inclusion of new
parameters.
Measures of QoE concerns user performance based on
actual usage. There are numerous ways to gather objective
QoE for laboratory and field tests or different types of
services and user scenarios. A key point relates objective
measures of QoE to user perception and beyond. This
takes QoE into the realm of user experience. However, in
the study of QoE, subjective user measure remains an
important variable. Thus, a combination of objective and
subjective QoE measurements better reflects the complex
nature of QoE.
III. QUALITY OF EXPERIENCE METRICS
In the context of HAS, TCP throughput reduces with
mishaps along the network path, for example, packet
losses and reordering. In such circumstances, video
playback pauses for new video data. This severely impacts
user-QoE. Many other factors affect user-QoE, such as
quality of video and smoothness of playback, [16]. A
Mean Opinion Score (MOS) of 1 ("Bad") to 5
("Excellent") expresses QoE, [8]. Subjective or objective
measures accumulates to mean opinion scores. However,
objective measures, such as Peak-Signal-to-Noise-Ratio
(PSNR) and Mean Square Error (MSE), are not suitable
for HTTP video streaming. They only evaluate spatial
video quality. Some application performance metrics
include, [37]:
 Initial buffering time: Tinit - measures the period
between starting time of loading a video and
starting time of playing it.
 Mean rebuffering duration: Trebuf - measures the
average duration of a rebuffering event.
 Rebuffering frequency: frebuf - measures
rebuffering event frequency. In some cases, the
amount of data in the video buffer decreases to
low values. Playback pauses. The player enters a
rebuffering state.
The development of QoE evaluation methodologies,
performance metrics and reporting protocols play a key
role in optimizing the delivery of HAS services, [18].
3GPP DASH defines the following QoE metrics, in
specification TS 26.247, [23], [32]:

3517
 HTTP request/response transactions: logs the
outcome of each HTTP request and corresponding
HTTP response. For every HTTP request/response
transaction, the player measures and reports:
o Type of request (e.g., MPD, initialization
segment, media segment)
o Time of receiving HTTP request and
corresponding HTTP response (in wall clock
time)
o HTTP response code
o Contents in the byte-range-spec part of HTTP
range header
o TCP connection identifier
o Throughput trace values for successful requests
from HTTP request/response transactions. It is
also possible to derive more specific
performance metrics, such as (a) the fetch
durations of the MPD, (b) initialization segment
and (c) media segments.
 Representation switch events: reports a list of
representation switch events taking place during a
measurement interval. A representation switch
event signals the player’s decision to perform a
representation switch to a new representation. As
part of each representation switch event, cf. Figure
3, the player reports: (a) the identifier for the new
representation, (b) time of the switch event (in
wall clock time) and (c) media time of the earliest
media sample played out from the new
representation.
Fig. 3. Different types of switching are recorded during
streaming.
 Average throughput: indicates the average
throughput the player observes during the
measurement interval. As part of the average
throughput metric, the player measures and
reports:
o Total number of content bytes (i.e., the total
number of bytes in the body of the HTTP
responses) received during the measurement
interval
o Activity time during the measurement interval:
time during which at least one GET request is
still not completed
o Wall clock time and duration of the
measurement interval
o Access bearer for the TCP connection based on
the average throughput report
o Type of inactivity (e.g., pause of presentation)
 Initial playout delay: signals initial playout delay
at the start of the streaming. Initial delay is always
present in adaptive video streaming services, as a
certain quantity of data must be transferred before
decoding and playback can begin. At first, the
video playback delays more than necessary in
order to fill the playout buffer. The playout buffer
is an efficient tool. It handles short term
throughput variations. However, the amount of
playtime it initially buffers trades-off with actual
length of the corresponding delay, [28]. Playout
buffer measures the time from when the player
requests the first media segment, to the time media
is retrieved from the player buffer.
 Buffer level: provides a list of buffer occupancy
level measurements carried out during playout, cf.
Figure 4. As part of the buffer level metric, the
player measures and reports the buffer level
indicating the playout duration for which media
data is available. This starts from the current
playout time along with the time for measurement
of the buffer level.
Fig. 4. Playout buffer fills with video segments.
 Play list: log a list of playback periods in the
measurement interval. Each playback period is the
time interval between a user action and the first
occurring among: (a) the next user action, (b) the
end of playback or (c) a failure that stops
playback. The type of user action triggering
playout may include: (a) a new playout request, (b)
resume playout from pause or (c) user-requested
quality change. For each playback period, the
player measures and reports: (a) the identifiers of
the representations, (b) rendering times (in media
time) and (c) rendering durations, (d) playback
speed relative to normal playback speed (e.g., to
track trick modes, such as fast forward or rewind)
and (e) reasons for interrupts of continuous
playback for this representation (e.g., due to
representation switch events, rebuffering, user
request, end of period, media content and metrics
collection period).

3518
Fig. 5. Adaptive video streaming file fragmentation. [17]
 MPD information: reports information on the
media presentations. It obtains information from
the MPD, so that servers, without direct access to
the MPD, can learn the media characteristics. The
player reports media representation attributes on
bit rate, resolution, quality ranking and codec-
related media information, (including profile and
level) utilizing this metric, cf. Figure 5.
Fig. 5. Some information an MPD file contains.
Stalling stops video playback because the playout buffer
underruns or becomes low/empty, [28]. During streaming
a case may occur, where the video streaming application
utilizes more data than incoming video bitrates. The
playout buffer becomes smaller. However, the video
eventually stops, when there is insufficient data in the
playout buffer, for playback to continue. The length of an
interruption guides re-buffering. Users experience longer
stalling durations, with large playtime buffering. Viewers
prefer a scenario with a single long freeze event, when
they compare it to one with frequent short freezes, [21].
Research in [7] observes overall video quality decreases as
duration of the impairment increased. Further, video stalls
are worse than frame rate reduction. They observe video
stalls at irregular intervals are worse than those at periodic
intervals. Bandwidth fluctuations can cause playout
interruptions of video. Packet loss through buffer overruns
(buffer is full, so receipt of packets results in drops
occurring) lead to retransmission of some packets and
consequent delays. Playout interruptions annoy users and
delays create bad effects, such as flickering. These events
should be taken into consideration for QoE estimation.
Current adaptive streaming approaches, [10] [4] [14],
focus on the following five QoE metrics, cf. Figure 6.
Fig. 6. Favorable user-perceived QoE for adaptive video
streaming.
IV. A TRANSPORT LAYER PROTOCOL: TCP
Applications requiring data packet delivery guarantees
utilize Transmission Control Protocol (TCP) as their
transport layer protocol, [20]. It is a sliding window
protocol, [11]. TCP handles timeouts and retransmissions.
TCP establishes a full duplex virtual connection between
two endpoints. An IP address and TCP port number
defines each endpoint. The operation of TCP as a finite
state machine is made possible in its implementation. The
server transfers byte stream segments across one or more
network channels. The window size determines the
number of data bytes sent, before an acknowledgment
(ACK) from the receiver is necessary. We discuss two
important algorithms TCP utilizes: (1) Congestion
avoidance and (2) Fast start. We select these as they have
huge impacts on video streaming. We explain how these
algorithms work. This would later enable an understanding
of TCP interactions with video flows during streaming.
A.Congestion Avoidance
The TCP Congestion Avoidance algorithm, [25], heavily
influences the behavior of Internet traffic dynamics.
Congestion occurs when the receiver consumes data at a
slower rate than the sender transmits. The buffers at the
receiver overflows. This results in lost packets, which the
receiver retransmits. Congestion occurs at a router, when
the capacity of the output link is less than the sum of all
inputs (assume multiple input links and one outgoing link).
Congestion avoidance deals with lost packets.
There are two ways of indicating that a packet was lost:
(1) a timeout occurs and (3) receipt of duplicate ACKs.
Congestion avoidance and slow start are two separate
algorithms, each with its own goal. If TCP detects
congestion, it slows down transmission packet rates into
the network. TCP then call on slow start to resume until
congestion occurs, again. Thus, in practice one
implements congestion avoidance and slow start together.
A connection maintains two variables, when combining
congestion avoidance and slow start: (1) a congestion
window, cwnd and (2) a slow start threshold size, ssthresh.
These algorithms operate as follows, [30]:
1. Initialization for a given connection sets cwnd to one
segment and ssthresh to 65535 bytes.
2. The TCP output routine never sends more than the
minimum of cwnd and window size the receiver
advertises.
3. As congestion occurs (a timeout or the reception of
duplicate ACKs), ssthresh saves one-half of the

3519
current window size (the minimum of cwnd and
receiver's window advertise size, but at least two
segments). Additionally, if a timeout indicates
congestion, cwnd is set to one segment (slow start).
4. As new data is acknowledged by the other end,
increase cwnd. However, the way it increases
depends on whether TCP is performing slow start or
congestion avoidance.
TCP is in slow start if the cwnd is less than or equal to
ssthresh. Otherwise, TCP is using the congestion
avoidance algorithm. The slow start algorithm continues
as long as TCP is less than half of the point. That is, the
previously set value, when congestion occurs. Then,
congestion avoidance commences.
The congestion avoidance algorithm increments the cwnd
by segment size times segment size divided by cwnd,
every time it receives an ACK, where cwnd is the
command window size in bytes. The growth rate is linear.
Thus, the increase in cwnd is at most one segment each
round-trip time. This occurs regardless of how many
ACKs the algorithm receives in the RTT. However, slow
start increases cwnd by the number of ACKs it receives in
a RTT. Figure 7 depicts the congestion avoidance
mechanism. Shown also is fast recovery or start, which is
discussed next.
Fig. 7. TCP Congestion avoidance mechanism. [24]
B.Fast Start
TCP fast start [19] speeds up short Web transfers. The
sender caches network parameters, which avoids paying a
slow start penalty for each download. Stale cache
information degrades performance. Fast start enables TCP
connections to reuse network information from the recent
past. Therefore, it does not repeat the slow start discovery
algorithm each time, especially ones transferring small
amounts of data between pauses. Cached information
includes:
 Congestion window size (cwnd)
 Slow start threshold (ssthresh)
 Smoothed round-trip time (srtt)
 Variance of the smoothed round-trip (rttvar)
Two important goals, [19] are:
 If the cached information is still valid, fast start
should help improve performance. However, if
this information is stale (for instance, because of a
sudden surge in network load), fast start should not
result in worse performance than if TCP utilizes
the standard slow start algorithm in the first place.
 The performance gains of fast start should not be
at the expense of other connections. It is okay for
other connections to suffer, once fast start tries to
use only its share of the bottleneck bandwidth.
But, not because the fast start connection is being
too aggressive.
It is very difficult to meet these objectives, without
increasing the network complexity, for example,
connection-to-connection state in routers. However, TCP
fast start does its best with a simple drop priority
mechanism, which routers implement.
We now present a simple example showing the growth
rate of packets sent over a channel using TCP fast start.
Initially, the sender transmits one segment and waits for an
ACK. Let us assume the ACK is sent successfully from
the receiver to the sender. When the sender receives an
ACK, it increments the congestion window one, that is,
from one to two. Then, two the sender transmits two
segments. Again, let us assume that the ACKs for these
two segments are sent successfully from the receiver to the
sender. The sender increases the ACK count to four and
the congestion window to four, see Figure 8 (slow start,
but increase rapidly so really fast start as labelled in the
figure). This process produces what seems to be an
exponential growth rate. However, the receiver may delay
its ACKs, generally sending one ACK for every two
segments it receives. Thus, it is not exactly exponential. In
addition, at a point in transmission, the network,
intermediate network device or router reaches their
maximum capacity. The intermediate network device
starts to discard packets. The sender now realizes the
congestion window is too large. A simple way to view
TCP: whenever there is congestion, the input rate
decreases, and whenever there is under-utilization input
rate increases.
Fig. 8. TCP Fast Start: increase by one packet per ACK

3520
V. EXPERIMENTS
The following experiments use the conventional DASH
controller and measures five DASH-based performance
metrics. All experiments are hosted on a Windows 10
machine, with the following specifications: Intel(R)
Core(TM) i7-5500U CPU 2.40GHz processor, 16.00 GB
physical memory and an Intel(R) HD Graphics processor.
A virtual network is setup on the Windows 10 machine for
the emulation test bed. Our setup consists of a video server
running Ubuntu (HTTP server), a router running FreeBSD
(Home Router) and two real network players (HAS client)
hosted on Ubuntu, cf. Figure 9.
Fig. 9. Experiment testbed setup.
Fig.10. Quality levels: Conventional DASH players.
The result of two conventional players competing at a
5Mbps bottleneck link is shown on Figure 10. The players
concurrently download the Elephant’s dream video [5]. As
shown the players are not able to maintain high quality
video and so the user experience degrades at certain time
intervals. The reason for this is discussed in section VI.
VI. DISCUSSION: TCP AND DASH
Generally, the sender controls the aggressiveness of TCP.
However, the receiver throttles traffic in DASH-based
systems, cf. Figure 11. TCP tries to increase bandwidth
utilization and avoid congestion. However, a DASH player
ensures no video playback interruptions by pre-fetching
and buffering sufficient video segments. Thus, the DASH
player continuously increases its receiver window during
ON periods to obtain as much video segments as possible,
[4]. The sender transmits as much traffic as possible in
bursts. Segments transmissions occur independently of
each other. This happens until the player’s playout buffer
has enough video segments, which either changes the
player to an OFF period or the sender incurs TCP packet
losses. The connection resets between ON-OFF periods,
[10]. This causes extreme bursty traffic (cf. Figure 12) and
TCP inefficiency, [38]. The result is unstable video
playback quality and unfair sharing of network resources.
VII. CONCLUSION
We introduce DASH-based video delivery. We show the
need for QoE in evaluating DASH-based systems. We
highlight the main QoE metrics researchers utilize in
evaluating DASH-based systems. We explore the working
mechanisms of TCP: (1) Congestion and (2) Fast Start.
We present an experiment with competing DASH players.
The experiment shows poor QoE for existing users.
Further, we explain how the two TCP mechanisms result
in this poor user-QoE for DASH users.
Fig.11. The player first makes a GET request for the MPD
file containing data about the video segments it wants to
download. After this initiation the player makes requests
for successive segments. This is on a segment by segment
basis and gives the player the option to select the next
segment based on certain "desired" criteria.

3521
Fig. 12. Bursty network traffic. [26]
REFERENCES
[1] Andrews, M., K. Kumaran, K. Ramanan, A. Stolyar,
P. Whiting, and R. Vijayakumar (2001). Providing
quality of service over a shared wireless link. IEEE
Communications magazine 39 (2), 150-154.
[2] Brooks, P. and B. Hestnes (2010). User measures of
quality of experience: why being objective and
quantitative is important. IEEE network 24 (2).
[3] Brooks, P. and B. Hestnes (2010). User measures of
quality of experience: why being objective and
quantitative is important. IEEE network 24 (2).
[4] De Cicco, L., V. Caldaralo, V. Palmisano, and S.
Mascolo (2013). Elastic: a client-side controller for
dynamic adaptive streaming over http (dash). In 2013
20th International Packet Video Workshop, pp. 1-8.
IEEE.
[5] De Cicco, Luca, and Saverio Mascolo. "An
experimental investigation of the Akamai adaptive
video streaming." In Symposium of the Austrian HCI
and Usability Engineering Group, pp. 447-464.
Springer, Berlin, Heidelberg, 2010.
[6] Ferguson, P. and G. Huston (1998). Quality of
service: delivering QoS on the Internet and in
corporate networks, Volume 1. Wiley New York.
[7] Huynh-Thu, Q. and M. Ghanbari (2008). Temporal
aspect of perceived quality in mobile video
broadcasting. IEEE Transactions on Broadcasting 54
(3), 641-651.
[8] ITU-T RECOMMENDATION, P. (1998). Subjective
audiovisual quality assessment methods for
multimedia applications.
[9] Jain, L. P., W. J. Scheirer, and T. E. Boult (2004).
Quality of experience. In IEEE multimedia. Citeseer.
[10]Jiang, J., V. Sekar, and H. Zhang (2012). Improving
fairness, efficiency, and stability in http-based
adaptive video streaming with festive. In Proceedings
of the 8th international conference on Emerging
networking experiments and technologies, pp. 97-108.
ACM.
[11]Kontaki, M., A. N. Papadopoulos, and Y.
Manolopoulos (2007). Adaptive similarity search in
streaming time series with sliding windows. Data &
Knowledge Engineering 63 (2), 478-502.
[12]Laghari, K. U. R. and K. Connelly (2012). Toward
total quality of experience: A qoe model in a
communication ecosystem. IEEE Communications
Magazine 50 (4).
[13]Lederer, S., C. Müller, and C. Timmerer (2012).
Dynamic adaptive streaming over http dataset. In
Proceedings of the 3rd Multimedia Systems
Conference, pp. 89-94. ACM.
[14]Li, Zhi, Xiaoqing Zhu, Joshua Gahm, Rong Pan, Hao
Hu, Ali C. Begen, and David Oran. "Probe and adapt:
Rate adaptation for HTTP video streaming at
scale." IEEE Journal on Selected Areas in
Communications 32, no. 4 (2014): 719-733.
[15]McDysan, D. E. (2000). QoS & tra-c management in
IP & ATM networks, Volume 18. McGraw-Hill New
York, NY, USA.
[16]Mok, R. K., E. W. Chan, and R. K. Chang (2011).
Measuring the quality of experience of http video
streaming. In Integrated Network Management (IM),
2011 IFIP/IEEE International Symposium on, pp.
485-492. IEEE.
[17]Nwamba, C. (2017). How to Implement Smooth
Video Buffering for a Better Viewing Experience
[Blog post]. Retrieved from
https://cloudinary.com/blog/how_to_implement_smoo
th_video_buffering_for_a_better_viewing_experience
.
[18]Oyman, O. and S. Singh (2012a). Quality of
experience for http adaptive streaming services. IEEE
Communications Magazine 50 (4).
[19]Padmanabhan, V. N. and R. H. Katz (1998). Tcp fast
start: A technique for speeding up web transfers.
[20]Postel, J. (1981). Transmission control protocol.
[21]Qi, Y. and M. Dai (2006). The effect of frame
freezing and frame skipping on video quality. In
Intelligent Information Hiding and Multimedia Signal
Processing, 2006. IIH-MSP'06. International
Conference on, pp. 423-426. IEEE.
[22]Recommendation, I.-T. (2001). Perceptual evaluation
of speech quality (pesq): An objective method for
end-to-end speech quality assessment of narrowband

3522
telephone networks and speech codecs. Rec. ITU-T P.
862.
[23]Reznik, Y., E. Asbun, Z. Chen, and R. Vanam (2013,
April 23). Method and apparatus for smooth stream
switching in mpeg/3gpp-dash. US Patent App.
13/868,968.
[24]Rhee, Injong, Volkan Ozdemir, and Yung Yi. TEAR:
TCP emulation at receivers-flow control for
multimedia streaming. NCSU Technical Report, 2000.
[25]Ryu, S., C. Rump, and C. Qiao (2003). Advances in
internet congestion control. IEEE Communications
Surveys & Tutorials 5 (1).
[26]Sarvotham, Shriram, Rudolf Riedi, and Richard
Baraniuk. "Connection-level analysis and modeling of
network traffic." In Proceedings of the 1st ACM
SIGCOMM Workshop on Internet Measurement, pp.
99-103. ACM, 2001.
[27]Schulzrinne, Henning. "Real time streaming protocol
(RTSP)." (1998).
[28]Seufert, M., S. Egger, M. Slanina, T. Zinner, T.
Hoÿfeld, and P. Tran-Gia (2015). A survey on quality
of experience of http adaptive streaming. IEEE
Communications Surveys and Tutorials 17 (1), 469-
492.
[29]Sodagar, I. (2011). The mpeg-dash standard for
multimedia streaming over the internet. IEEE
MultiMedia 18 (4), 62-67.
[30]Stevens, W. R. (1997). Tcp slow start, congestion
avoidance, fast retransmit, and fast recovery
algorithms.
[31]Stockhammer, T. (2011a). Dynamic adaptive
streaming over http: standards and design principles.
In Proceedings of the second annual ACM conference
on Multimedia systems, pp. 133-144. ACM.
[32]Stockhammer, T. (2011b). Ts 26.247 transparent end-
to-end packet-switched streaming service (pss).
Progressive Download and Dynamic Adaptive
Streaming over HTTP, 3GPP.
[33]Takahashi, A., D. Hands, and V. Barriac (2008).
Standardization activities in the itu for a qoe
assessment of iptv. IEEE Communications Magazine
46 (2).
[34]Tavakoli, S., J. Gutiérrez, and N. Garcia (2014).
Subjective quality study of adaptive streaming of
monoscopic and stereoscopic video. IEEE Journal on
Selected Areas in Communications 32 (4), 684-692.
[35]Wang, Z. and J. Crowcroft (1996). Quality-of-service
routing for supporting multimedia applications. IEEE
Journal on selected areas in communications 14 (7),
1228_1234.
[36]Wang, Z., S. Banerjee, and S. Jamin (2003). Studying
streaming video quality: from an application point of
view. In Proceedings of the eleventh ACM
international conference on Multimedia, pp. 327-330.
ACM.
[37]Wang, Z., S. Banerjee, and S. Jamin (2003). Studying
streaming video quality: from an application point of
view. In Proceedings of the eleventh ACM
international conference on Multimedia, pp. 327-330.
ACM.
[38]Zhou, L., X. Wang, W. Tu, G.-M. Muntean, and B.
Geller (2010). Distributed scheduling scheme for
video streaming over multi-channel multi-radio
multihop wireless networks. IEEE Journal on Selected
Areas in Communications 28 (3).
Biographies and Photographs
Koffka Khan received the M.Sc., and
M.Phil. degrees from the University of the
West Indies. He is currently a PhD student
and has up-to-date, published numerous
papers in journals & proceedings of international repute.
His research areas are computational intelligence, routing
protocols, wireless communications, information security
and adaptive streaming controllers.
Wayne Goodridge is a Lecturer in the
Department of Computing and Information
Technology, The University of the West
Indies, St. Augustine. He did is PhD at
Dalhousie University and his research interest
includes computer communications and security.

QoE in DASH

More Related Content

What's hot (18)

Similar to QoE in DASH (20)

More from Eswar Publications (20)

Recently uploaded (20)

QoE in DASH