Key Research Issues Of Quality Of Experience:Measurement Of In-Service QoE

Quality of Experience (QoE) is the perceived quality of any network ser-vice or product by the end users and is a measure of user satisfaction based on objective network related parameters and subjective user related parameters. QoE deals with the end to end quality of networks and services. It is different from Quality of Service (QoS) as QoS is more network oriented and deals with the network parameters like delay, packet loss and jitter etc. whereas QoE is more user centric and deals with the user perception and satisfaction. Previous-ly QoE is either measured subjectively based on the user mean opinion scale (MOS) ratings or objectively based on user perceived media quality models. But these approaches are not holistic and dynamic in nature. Hence the key research issue is to develop QoE approach which can be used to measure the real time QoE quantitatively based on subjective as well as objective approach. This approach must be implemented to measure the in-service QoE dynamical-ly i.e. the change in the situation (change in the environment, network, service and user behavior parameters at the same time) must be reflected in the mea-sured QoE. We have proposed a novel context aware and user centric approach where the subjective part of QoE is calculated using the user behavior model- ing parameters and the objective part is measured using the context awareness parameters. The subjective part of our approach is based on user perception and other behavior parameters whereas the objective part is based on network, environment, device and other contextual parameters. The next problem is to implement this approach so that the measured QoE is dynamic in nature and accurate.We have used Bayesian Networks to implement our proposed approach and measure the estimated QoE for P2P streaming systems. This is an esti-mation and prediction methodology based on relationships of nodes and con-ditional probability tables. We have implemented our approach for both P2P live streaming and P2P Video on Demand (VoD) streaming systems. Our goal is to estimate and predict the in-service QoE. The basic task of the proposed Bayesian network is to update and compute the posterior probability distribu-tion for a set of query variables also known as beliefs, given some assignment of values to a set of evidence variables. We have proposed the nodes of the Bayesian network, the relationships between these nodes, the conditional prob-ability tables for each node, Bayesian network structure, value nodes and the mapping of calculated value on mean opinion score (MOS) scale to get the final scalar QoE value. We then validated our results against different objective and subjective models.We have used Decision Networks to implement our proposed approach and measure the estimated QoE for P2P streaming systems. We have implemented our approach for both P2P live streaming and P2P VoD streaming systems. De-cision network model combine Bayesian networks with additional node types for actions and utilities. QoE is estimated based on the evidence variables when available to the chance nodes. We have proposed the nodes of the decision net-work, the relationships between these nodes, the conditional probability tables for each node, decision network structure, value nodes and decision node which calculate the final QoE based on the available choices. We then validated our results against different objective and subjective models.We have implemented both our proposed methods i.e. Bayesian Network-s and Decision Networks for P2P live video streaming systems and P2P VoD video streaming systems under similar conditions. Based on the results, both the methods are compared with each other. Bayesian Network methodology is more accurate in dynamically measuring the change in QoE and giving the utility values of user, network and environment context. Hence it gives a more useful information regarding the problem areas in case of poor QoE. While De-cision Network methodology is more appropriate for autonomic service frame-work where system dynamically adapt to the changing environment. Also a QoE management framework has been proposed for future work.The results show that the estimated QoE is not only user centric and context aware but also holistic and dynamic in nature. Any change in the environment, network, service or user parameters are reflected in the final QoE. This thesis essentially proposes and implements a QoE measurement approach which can help improve existing services, generate new services, implement better net-work architecture designs and provide help to content providers for better user satisfaction.