Transcript qqa - Irisa

QQA: Quantitative
Quality Assesment
(or pseudo-subjective quality)
in @rmor’s evaluation,
22-23 october 2003
Global view

QQA: Quantitative Quality
Assessment
(or pseudo-subjective quality
assessment)



Quantitative evaluation of
quality

as perceived by the
observer,

automatically, and, if
necessary, in real time.
Idea: to use specific learning
tools (particular open queuing
networks) capturing the way
human react, taking measurable
quantities as inputs.
Objective reached. Tested on:
video and audio separately.
Applications under analysis:
control; monitoring.
Extensions under analysis:
to pricing, to diffserv architectures, to
traffic prediction and bandwidth
negotiation, to control issues in radio
access networks, to home networking.
@rmor – INRIA Rennes
The method




Use a particularly performant statistical learning tool:
a product form queueing network with positive and negative
customers (a G-network, or RNN)
to learn how humans react face to a multimedia stream
after having passed through a packet network.
Key points:




identify appropriate input variables (loss rate, source bit rate, …)
a configuration = a set of values for the input variables
with each configuration associate a quality value given by a set of real
observers under controlled conditions
find a G-network with the mapping:



input variables = external arrival rates (of positive customers)
only one queue sends customers outside, and the quality is mapped to the
load of this node
for each configuration, the load of the exit node is (very) close to the quality
given by the human observers
@rmor – INRIA Rennes
Example of implementation for video
Source
IP network
stream of
voice,
music,
video,
multimedia,
…
Receiver
RNN
asking the source
for BR, FR, RA
@rmor – INRIA Rennes
measuring
LR, CLP
For the remaining extensions:
People

G. Rubino, DR INRIA

S. Mohamed,
PhD (January 2003),
now temporary engineer
M. Varela, PhD student
(starting his 2nd year)



B. Tuffin, CR INRIA,
Y. Hezel, PhD student,
for pricing issues

J.-M. Bonnin, MdC ENST B,
for mobile applications

D. Ros, MdC ENST B,
J. Orozco, PhD student,
for control in diffserv

L. Toutain, MdC ENST B,
S. Ben Hamida, PhD student,
for control in home networking
(conditional to STREP accepted)
F. Cervantes, J. Incera, prof. at
ITAM, Mexico, for dynamic
bandwidth negotiation
@rmor – INRIA Rennes
For the remaining extensions:
People

G. Rubino, DR INRIA

S. Mohamed,
PhD (January 2003),
now temporary engineer
M. Varela, PhD student
(starting his 2nd year)



B. Tuffin, CR INRIA,
Y. Hezel, PhD student,
for pricing issues

J.-M. Bonnin, MdC ENST B,
for mobile applications

D. Ros, MdC ENST B,
J. Orozco, PhD student,
for control in diffserv

L. Toutain, MdC ENST B,
S. Ben Hamida, PhD student,
for control in home networking
(conditional to STREP accepted)
F. Cervantes, J. Incera, prof. at
ITAM, Mexico, for dynamic
bandwidth negotiation
@rmor – INRIA Rennes
For the remaining extensions:
People

G. Rubino, DR INRIA

S. Mohamed,
PhD (January 2003),
now temporary engineer
M. Varela, PhD student
(starting his 2nd year)



B. Tuffin, CR INRIA,
Y. Hezel, PhD student,
for pricing issues

J.-M. Bonnin, MdC ENST B,
for mobile applications

D. Ros, MdC ENST B,
J. Orozco, PhD student,
for control in diffserv

L. Toutain, MdC ENST B,
S. Ben Hamida, PhD student,
for control in home networking
(conditional to STREP accepted)
F. Cervantes, J. Incera, prof. at
ITAM, Mexico, for dynamic
bandwidth negotiation
@rmor – INRIA Rennes
For the remaining extensions:
People

G. Rubino, DR INRIA

S. Mohamed,
PhD (January 2003),
now temporary engineer
M. Varela, PhD student
(starting his 2nd year)



B. Tuffin, CR INRIA,
Y. Hayel, PhD student,
for pricing issues

J.-M. Bonnin, MdC ENST B,
for mobile applications

D. Ros, MdC ENST B,
J. Orozco, PhD student,
for control in diffserv

L. Toutain, MdC ENST B,
S. Ben Hamida, PhD student,
for control in home networking
(conditional to STREP accepted)
F. Cervantes, J. Incera, prof. at
ITAM, Mexico, for dynamic
bandwidth negotiation
@rmor – INRIA Rennes
Publications



“A Study of Real--time Packet Video Quality Using Random Neural
Networks”.
S. Mohamed and G. Rubino. IEEE Transactions On Circuits and Systems
for Video Technology, Vol. 12, No. 12, December 2002.
“Performance evaluation of real-time speech through a packet
network: a Random Neural Networks based approach”.
S. Mohamed, G. Rubino and M. Varela.
To appear in Performance Evaluation.
Other publications in



Infocom 2001
ICOIN’15, 2001
PDPTA’2001
@rmor – INRIA Rennes
Next future


develop a video-conference tool with automatic quality control
based on QQA
transform the approach into an industrial product


extend the approach in coupling traffic prediction with dynamic
negotiation of bandwidth





idea: put a dynamic bandwidth negotiator at the edge of the core
use QQA and traffic prediction (+ a pricing scheme) to allow the user to
negotiate with the provider
apply QQA to control in


Phillips? France Telecom?
a diffeserv architecture
a home network (together with reservation techniques, network
calculus tools and IPv6 facilities)
in pricing (to build virtual user profiles);
to explore the interest of the same tools in risk evaluation, and in
compression techniques
@rmor – INRIA Rennes
10
Next future: on the tool

improve the mathematical analysis in the case of
recurrent networks


and then, apply it to the WAN design area
improve the numerical algorithms used to analyze the
networks

basically, by adapting to G-networks specific techniques that
have proven to be efficient with ANN
@rmor – INRIA Rennes
11