Overview of the Dissertation

In this Chapter, we have introduced the motivations and the context of the problems that we tackle in this dissertation. The rest of the report is divided into three parts. In the first one, we focus on the automatic multimedia quality assessment problem. In Chapter 3, a state-of-the-art of the issues related to our work is presented. More specifically, we discuss the current technology concerning objective multimedia quality measures, real-time transmission of multimedia streams over packet networks, the challenges and the existing methods to overcome them, audio and video compression and several standard codecs, and neural networks are overviewed. In Chapter 4, we present a new method to measure real-time multimedia quality when it is transmitted in real time over packet network. We provide the general methodology as it can be applied for speech, audio and/or video. We present also the most used techniques to evaluate subjective multimedia quality. We emphasize on the use of neural networks and we compare both ANN and RNN in the context of our problems. We describe also some possible uses and applications of our method. In the second part, we focus on speech and video quality as well as our rate control mechanism. In Chapter 5, we evaluate the applicability of our technique to evaluate, in real time, real-time speech transmitted over packet-based networks. We present the results of real experiments we carried out between national-wide and international-wide sites. The purpose of these experiments is to identify the most typical ranges of the main network parameters. We use three different speech encoders, namely PCM, ADPCM and GSM. Three different spoken languages (Arabic, Spanish and French) are considered. Other important parameters we take into account are loss rate, consecutive loss duration and the packetization interval of the speech. In Chapter 6, we evaluate the applicability of our technique to evaluate real-time video transmitted over packet-based networks in real time. We use the H.263 encoder and we selected five important parameters that have a priori an impact on video quality. These parameters are bit rate, frame rate, loss rate, consecutive loss duration, and the amount of redundant information to protect against loss. In Chapter 7, we study the impact of quality-affecting parameters on the perceived quality. Our study is for both video and audio/speech applications. This study can help in understanding the behavior of both packet video and audio in the Internet (or in any packet-based network). In Chapter 8, we propose a rate control protocol that takes into account the end-user perception as well as the traditional network measurements to deliver the best possible quality in real time given the network current state. We provide a study of the possible parameters that can be used for the both video and audio applications. The new protocol can, in certain situations, save network resources while maintaining TCP-friendness behavior. The third and last part of this report deals with traffic prediction and random neural network learning algorithms. In Chapter 9, we provide a new method for traffic prediction that takes into account not only the short term dependency of the traffic process, but also the long term. We compare our approach with the existing ones and we evaluate it using real traffic traces. In Chapter 10, we present two new training algorithms to improve the performance and the accuracy of RNN. We evaluate these algorithms and we also compare them with the traditional gradient descent algorithm. Finally, we give the conclusions of this dissertation and some future research directions related to the problems addressed during this work.