Content uploaded by Sami Souihi
All content in this area was uploaded by Sami Souihi on Feb 21, 2017
Content may be subject to copyright.
An Open Source Platform for Perceived Video Quality
Lamine Amour, Sami Souihi, Said Hoceini and Abdelhamid Mellouk
University of Paris-Est Créteil Val de Marne (UPEC)
Image, Signal and Intelligent Systems Lab-LiSSi and Netw & Telecoms Dept, IUT CV
122 rue Paul Armangot, 94400 Vitry sur Seine, France
(lamine.amour, sami.souihi, hoceini, mellouk) @u-pec.fr
To ensure the best multimedia service quality in order to well
address users’ expectations, a new concept named Quality of
Experience (QoE) has appeared. Two methods can be used to
evaluate the user satisfaction, a subjective one and an objective
one. The subjective approach is based on measured real data.
The problem is there is no dataset large enough and can be
used to well evaluate the QoE. In this, work we present our
approach to build a data set for subjective evaluation based
on a categorization approach and open source software.
Quality of Experience (QoE); Controlled environment; Crowd-
sourcing; Mean opinion score; Video; Framework.
A number of works are already made on the QoE area. For
example, Qualinet tries to quantify and propose eﬃcient QoE
estimation models. To build a more accurate model, the ﬁrst
need is to build a consistent database. User perception can be
is inﬂuenced by a huge number of parameters (User Proﬁle,
Network parameters, Application parameters, ...). These fac-
tors are called QoE Inﬂuence Factors (QoE IFs) . Nowadays,
there is no available database large enough to includes all these
QoE IFs and can be used to produce a better accurate esti-
mation model. In this context, we present CLLF (Controlled
LiSS i −Lab F ramewor k) as an open source platform to help
researchers to build a large QoE/QoE IFs database for video
streaming services (Youtube).
2. RELATED WORK
The QoE is a hot topic given the huge number of works that
we can ﬁnd in the literature. In this section, we will present
two video QoE frameworks.
- The ﬁrst framework is proposed by Figuerola Salas et al..
This system is used on a large scale with preliminary results of
a validation study. In fact, this system is based on an HTML5
Web-based tool that collects ratings of videos encoded at dif-
ferent bitrates. This work focuses on High Deﬁnition (720p
Permission to make digital or hard copies of part or all of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for proﬁt or commercial advantage and that copies
bear this notice and the full citation on the ﬁrst page. Copyrights for third-
party components of this work must be honored. For all other uses, contact
the Owner/Author(s). Copyright is held by the owner/author(s).
Q2SWinet’15, November 2–6, 2015, Cancun, Mexico.
(HD)) video service commonly employed for video services over
the Internet. In this framework, authors used diﬀerent quality
of video which are encoded from 784 Kbps and 1 Mbps with
AV C/H.264 codec, and used Mean Opinion Score method to
record the user’s perception.
- The second system is proposed by Hanhart et al.. In
this framework, the authors investigate two approaches to as-
sess a multi-view video depth (MVD) content on 2Ddisplays :
at ﬁrst by using a virtual view. Secondly, by using a free-
viewpoint video, which corresponds to a smooth camera mo-
tion during a time freeze. In fact, they conducted the crowd-
sourcing experiments using seven MVD sequences encoded at
diﬀerent bit rates and tested in both Lab-based evaluation and
Crowd-based evaluation, In this work, the single-stimulus (SS)
methodology was chosen as this methodology with the MOS
3. FRAMEWORKS DESCRIPTION
The main objective of our proposed platform  is to evalu-
ate the video streaming services user perception in controlled
environment which is achieved based on our previous proposal :
’QoE IFs hierarchical classiﬁcation’ made in . Based on QoE
IFs categories (Network, User proﬁle, Application, Device and
User feedback -Figure 1-), we will explain how to proceed to
built a framework with a lot of QoE IFs in a controlled envi-
Figure 1: QoE IFs categories presentation.
In fact, the platform goals are summarized in following ele-
ments : (i) An application has been developed and installed to
evaluate the video streaming services in controlled experimen-
tation. (ii) A platforme has been carried out, to implement our
architecture . (iii) Several codecs were considered ( 144pto
1080p(Full HD)) and several types of videos were used (e.g.
sport, movie, documentary, news, music, etc.) in the experi-
ments. In the following section, we will explain how we proceed
to collect QoE IFs and to achieve these goals.
3.2 How to collect QoE IFs categories ?
To build the dataset, we made a testbed in totally a con-
trolled environment. The Figure 2 presents the overall system
In this testbed, users give their MOS at the end of each
video. To consider all the QoE IFs that can have an im-
Figure 2: System components
pact on the user’s perception, we use and combine a lot of
softwares such as : VideoLAN Client player (VLC), Netem
(Network Emulator), Python language and MySQL database
server. Then, below we will describe how we proceed to collect
QoE IFs for each categorie (Figure 1).
3.2.1 UF and UP
User Feedback (UF) is a score given by user to subjectively
evaluate a service (MOS). It is represented by the bellow ques-
tions. The answer of each question is between 1 and 5 where
1)Starting video time : 1 : very long / 5 : very fast.
2)Lag between image and audio : 1 : very big lag/
5 : No lag
3)Image quality : 1 : very bad/ 5 : very good.
4)Sound quality : 1 : very bad/ 5 : very good.
5)MOS : 1 : very bad/ 5 : very good.
1 indicates that the quality is not acceptable or very bad (time
to start very long, lag between picture and audio very high,
etc.) and 5 indicates that the quality is very good (no buﬀer-
ing time, no lag, etc.). In addition, the platform implements
an ergonomic interface to ask users about their proﬁles (Age,
Sex, Study level and Experience with the video service).
Another important aspect is the device quality (QoD). The
proposed platform can collect devices information as the screen
size and the screen resolution, CPU performance, available
To evaluate the impact of network QoS, the platform uses
a network emulator called N etEm. A preliminary study was
done to help us to ﬁx the QoS variations value in order to get
out the combinations of QoS that show the problems in the
video stream as : video blocking, failure launch, lag between
the sound and image,.... In this experimentation, we test 3
QoS factors (delay, loss and rate.) [5, 6] and 117 combinations
summarized in the Table 1. These 3 QoS factors are varied in
a random manner.
QoE IFs Values
delay 0, 100, 200, 400 (ms)
rate 256, 512, 768, 1024, 1536 (kbitsrs)
loss 0, 5, 10 (%)
Table 1: QoS variations on the controlled laboratory
To collect the category application factors for the CLLF
framework, we worked on two sides. The ﬁrst part involves
the harvesting of traditional application parameters (e.g. bi-
trate, frame rate, etc). Our platform uses a modiﬁed version
of VideoLAN Client player (VLC) to gather all these QoE IFs.
The second part is the video content (e.g. video type, degree
of motion and codec). To implement the second part, (i) we
started with the selection of 8 video types based on existing
work. (ii) Once the choice is done, we downloaded 24 videos
with free rights (Youtube Creative Common) and with diﬀer-
ent resolutions (between 144pand 1080p), (iii) selected portion
of 30 seconds with a speciﬁc motion degree, (iv) uploaded them
an other time to Youtube and (v) call them with their URL
from the player . Figure 3 presents screen shots of the used
Figure 3: Screen shots of used videos.
Quality of Experience (QoE) appeared to improve network
control taking into account the real user’s perceived quality.
To attempt this goal, we have studied the QoE factors that
can impact the user’s satisfaction in a mobile controlled envi-
ronment in order to implement a platform that can be used to
build consistent and large databases. This latter can help us
to improve QoE estimation models. Our open source platform
is still in development to continue to introduce new QoE IFs
and enlarge dataset.
 L. Amour, S. Souihi, S. Hoceini, and A. Mellouk.
Platform video :. https://youtu.be/pMvfVQYplVk, 2015.
 L. Amour, S. Souihi, S. Hoceini, and A. Mellouk. A
hierarchical classiﬁcation model of qoe inﬂuence factors.
13th International Conference on Wired and Wireless
Internet Communications, May 25-27, 2015.
 O. Figuerola Salas, V. Adzic, A. Shah, and H. Kalva.
Assessing internet video quality using crowdsourcing. In
Proceedings of the 2Nd ACM International Workshop on
Crowdsourcing for Multimedia, CrowdMM ’13, pages
23–28, New York, NY, USA, 2013. ACM.
 P. Hanhart, P. Korshunov, and T. Ebrahimi.
Crowd-based quality assessment of multiview video plus
depth coding. 2014.
 A. Khan, L. Sun, and E. Ifeachor. Content clustering
based video quality prediction model for mpeg4 video
streaming over wireless networks. In Communications,
2009. ICC ’09. IEEE International Conference on, pages
1–5, June 2009.
 M. Mushtaq, B. Augustin, and A. Mellouk. Empirical
study based on machine learning approach to assess the
qos/qoe correlation. In Networks and Optical
Communications (NOC), 2012 17th European Conference
on, pages 1–7, June 2012.