Team dionysos

Overall Objectives
Scientific Foundations
Application Domains
New Results
Contracts and Grants with Industry
Other Grants and Activities

Section: New Results

QoE (Quality of Experience)

Participants : Gerardo Rubino, Kamal Singh.

We continue the development of the PSQA technology (Pseudo-Subjective Quality Assessment) in the area of Quality of Experience (QoE). PSQA is a method to build measuring modules capable of quantifying the quality of a video or an audio sequence, as perceived by the user, when received through an IP network. PSQA provides an accurate and efficiently computed evaluation of the quality. Accuracy means that PSQA gives values close to those than can be obtained from a panel of human observers, under a controlled subjective testing experiment, following an appropriate norm (which depends on the type of sequence or application). Efficiency means that our measuring tool can work in real time, if necessary. Observe that perceived quality is the main component of QoE.

At the heart of the PSQA approach there is the statistical learning process necessary to develop measuring modules. So far we have been using Random Neural Networks as our learning tool, where the specific learning technique was a classical Gradient Descent one. In [10] , we explored the use of much more efficient procedures, and in particular, we showed that the Levenberg-Marquardt method, well known in the standard Artificial Neural Network field, can be adapted to also improve significantly the performance of the basic learning algorithms. We also explored one of the most important variations of that algorithm, the so called “adaptive momentum” one, which works with RNN with the same efficiency as with ANN.

In [34] we followed the PSQA approach in the case of voice flows, but instead of learning from humans, we used as our reference the values given by PESQ to the sequences. PESQ is an automatic perceptual quality tool, belonging to the full reference family. In other words, PESQ computes a sort of distance between the original and the received sequences. It gives pretty good results but it can not work in real time, by definition. Our paper [34] is thus a first attempt to have a PESQ-like measuring tool working with no need to any reference, with promising results.

Since PSQA provides feedback in real time, one of the first applications area that come in mind is control. If we have a device capable of sending back to some deciding point a numerical evaluation of the perceived quality of a video or audio (or video and audio) communication, and if there is some action we can take to modify the system's state and thus, to possibly modify the quality, then there is an open way to optimize the ultimate target , the perception users have about the application / service they are using. This issues were the object of [28] where the main characteristics of PSQA for controlling purposes were discussed. See also  6.14 and  7.5 in this report.

In [29] (see also ) the present characteristics and performance of PSQA were described. In [63] , [27] , [65] , global presentations of our technology were given at different forums.


Logo Inria