Bessem Sayadi

QoE-based transport optimization for video delivery over next generation cellular networks

Video streaming is considered as one of the most important and challenging applications for next generation cellular networks. Current infrastructures are not prepared to deal with the increasing amount of video traffic. The current Internet, and in particular the mobile Internet, was not designed with video requirements in mind and, as a consequence, its architecture is very inefficient for handling video traffic. Enhancements are needed to cater for improved Quality of Experience (QoE) and improved reliability in a mobile network. In this paper we design a novel dynamic transport architecture for next generation mobile networks adapted to video service requirements. Its main novelty is the transport optimization of video delivery that is achieved through a QoE oriented redesign of networking mechanisms as well as the integration of Content Delivery Networks (CDN) techniques.

Superfluidity: a flexible functional architecture for 5G networks

We propose the innovative architecture of Superfluidity, a Horizon 2020 project, co-funded by the European Union. Superfluidity targets 5G networks, by addressing key network operator challenges with a multi-pronged approach, based on the concept of a flexible, highly adaptive, superfluid network. Superfluidity supports rapid service deployment and migration in a heterogeneous network environment, regardless of the underlying hardware. The overall proposal offers advanced capabilities in terms of service deployment and interoperability, while guaranteeing high-performance levels end-to-end. Copyright

Quality-fair HTTP adaptive streaming over LTE network

In HTTP adaptive streaming (HAS) applications multiple video clients sharing the same wireless channel may experience different video qualities as result of both different video content complexity and different channel conditions. This causes unfairness in the end-user video quality. In this paper, we propose a quality-fair adaptive streaming (QFAS) solution to deliver fair video quality to HAS clients competing for the same resources in an LTE cell. In the QFAS framework the share of radio resource is optimized according to video content characteristics and channel condition. The proposed solution is compared with other state-of-the-art strategies and numerical results in terms of SSIM quality metric shows that it significantly improves the quality fairness among heterogeneous HAS users.

Improving QoE and Fairness in HTTP Adaptive Streaming Over LTE Network

HTTP adaptive streaming (HAS) has emerged as the main technology for video streaming applications. Multiple HAS video clients sharing the same wireless channel may experience different video qualities as well as different play-out buffer levels, as a result of both different video content complexities and different channel conditions. This causes unfairness in the end-user quality of experience. In this paper, we propose a quality-fair adaptive streaming solution with fair buffer (QFAS-FB) to deliver fair video quality and to achieve asymptotically fair play-out buffer levels among HAS clients competing for the same wireless resources in a long-term evolution (LTE) cell. In the QFAS-FB framework, the share of radio resources is optimized according to video content characteristics, play-out buffer levels, and channel conditions. The proposed solution is compared with the other state-of-the-art strategies, and the numerical results show that it significantly improves the quality fairness among heterogeneous HAS users, reduces the video quality variations, and improves the fairness among the users play-out buffers.

Online learning for QoE-based video streaming to mobile receivers

This paper proposes a cross-layer control mechanism to stream efficiently scalable videos to mobile receivers. Its goal is to maximize the quality of the received video while accounting for the variations of the characteristics of the transmitted content and of the channel. The control problem is cast in the framework of Markov Decision Processes. The optimal actions to apply to the system are learned using reinforcement learning. For that purpose, the quality of the decoded frames at receiver is inferred by an observation (i) of the quality of the various scalability layers and (ii) of the level of queues at the Application and Medium Access Control layers of the transmitter only. Delayed as well as absence of information on the channel state are considered. Experiments show that the performance of the proposed solution is only slightly degraded with delayed or missing channel state information. The performance degradation is larger when considering a basic bitstream extractor, which serves as reference1.

Control of distributed servers for quality-fair delivery of multiple video streams

This paper proposes a quality-fair video delivery system able to transmit several encoded video streams to mobile users sharing some wireless resource. Video quality fairness, as well as similar delivery delay is targeted among streams. The proposed control system is implemented within some aggregator located near the bottleneck of the network. This is done by allocating the transmission rate among streams based on the quality of the already encoded and buffered packets in the aggregator. Encoding rate targets are evaluated by the aggregator and fed back to each remote video server, or directly evaluated by each server in a distributed way. Each encoding rate target is adjusted for each stream independently based on the corresponding buffering delay in the aggregator. The transmission and encoding rate control problems are addressed with a control-theoretic perspective. The system is described with a multi-input multi-output model and several Proportional Integral (PI) controllers are used to adjust the video quality as well as the buffering delay. The study of the system equilibrium and stability provides guidelines for choosing the parameters of the PI controllers. Experimental results show that better quality fairness is obtained compared to classical transmission rate fair streaming solutions while keeping similar buffering delays.

Joint Encoder and Buffer Control for Statistical Multiplexing of Multimedia Contents

Statistical multiplexing aims at transmitting several variable bit rate (VBR) encoded video streams over a bandlimited channel. Rate-distortion (RD) models for the encoded streams are often used to control the video encoders. As discrepancies frequently occur between the actual RD characteristics and their models, buffers are placed at the output of each coder to facilitate regulation. In this paper, a statistical multiplexer is proposed where video coders and buffers are controlled in a closed loop. First, a predictive joint rate controller accounting for maximum distortion, fairness, and smoothness constraints is considered. Second, all buffers are controlled simultaneously to limit deviations from a reference buffer occupancy to prevent buffer under and overflow. The main idea is to update the encoding rate for each encoding unit according to the average level of the buffers, to maximize the quality of each program. Simulation results show that the proposed scheme yields a smooth and fair video quality among programs thanks to the predictive control and allows an efficient use of the available bandwidth.

End-to-end stochastic scheduling of scalable video overtime-varying channels

This paper addresses the problem of video on demand delivery over a time-varying wireless channel. Packet scheduling and buffer management are jointly considered for scalable video transmission to adapt to the changing channel conditions. A proxy-based filtering algorithm among scalable layers is considered to maximize the decoded video quality at the receiver side while keeping a minimum playback margin. This problem is cast in the context of Markov Decision Processes which allows the design of foresighted policies maximizing some long-term reward. Experimental results illustrate the benefit of this approach compared to a shortterm policy in term of average PSNR improvement.

Analyzing the combination of different approaches for video transport optimization for next generation cellular networks

Various technologies for optimizing video delivery in cellular networks have been presented in the past. QoE-based transport optimizations attempt to look for optimal solutions to provide the best overall user experience for a group of users to leverage constrained network bandwidth. However, the design of these optimizations varies widely. An attractive idea is to deploy these optimizations together and expect more gains. In this article, we describe the interworking of several optimization approaches to analyze the feasibility of one single framework combining different approaches. We provide a first design of the joint framework that best leverages the advantage of each optimization approach.

Complexity comparison of the use of Vandermonde versus Hankel matrices to build systematic MDS Reed-Solomon codes

Reed Solomon RS(n, k) codes are Maximum Distance Separable (MDS) ideal codes that can be put into a systematic form, which makes them well suited to many situations. In this work we consider use-cases that rely on a software RS codec and for which the code is not fixed. This means that the application potentially uses a different RS(n, k) code each time, and this code needs to be built dynamically. A lightweight code creation scheme is therefore highly desirable, otherwise this stage would negatively impact the encoding and decoding times. Constructing such an RS code is equivalent to constructing its systematic generator matrix. Using the classic Vandermonde matrix approach to that purpose is feasible but adds significant complexity. In this paper we propose an alternative solution, based on Hankel matrices as the base matrix. We prove theoretically and experimentally that the code construction time and the number of operations performed to build the target RS code are largely in favor of the Hankel approach, which can be between 3.5 to 157 times faster than the Vandermonde approach, depending on the (n, k) parameters.