Cognitive Real-Time QoE Management in Video Streaming Services

Abstract

Satisfying the needs of users of online video streaming services requires not only to manage the network Quality of Service (QoS), but also to address the user's Quality of Experience (QoE) expectations. While QoS factors reflect the status of individual networks, they do not comprehensively capture the end-to-end features affecting the quality delivered to the user. In this situation, QoE management is the better option. However, traditionally used QoE management models require human interaction and have stringent requirements in terms of time and complexity. Thus, they fail to achieve successful performance in terms of real-timeliness, accuracy, scalability and adaptability. This dissertation work investigates new methods to bring QoE management to the level required by the real-time management of video services. In this paper, we highlight our main contributions. First, with the aim to perform a combined network-service assessment, we designed an experimental methodology able to map network QoS onto service QoE. Our methodology is meant to provide service and network providers with the means to pinpoint the working boundaries of their video-sets and to predict the effect of network policies on perception. Second, we developed a generic machine learning framework that allows deriving accurate predictive No Reference (NR) assessment metrics, based on simplistic NR QoE methods, that are functionally and computationally viable for real-time QoE evaluation. The tools, methods and conclusions derived from this dissertation conform a solid contribution to QoE management of video streaming services, opening new venues for further research.