FERRES I PRATS, J GARCÍA MATILLA, A AGUADED GÓMEZ, JI FERNANDEZ CAVIA, J FIGUERAS MAGDA BLANES, M (2011) “Competencia Mediática.
COMUNICACIÓN, EDUCACIÓN Y SOCIEDAD EN EL CONTEXTO DIGITAL
The measurement models that are proposed and analysed in this chapter are related to the emerging technologies for web services. It is clear that in this area the availability of acceptable objective performance metrics that correlate well with subjective scores are still in their early stages of development. Thus, more effect and standards are needed for defining a measurement model to represent the perceived quality which is experienced by end users commonly referred to as QoE.
Obviously, any requirement for a comprehensive QoE model should incorporate the merits of previous methods in order to have a suitable QoE model for packet traffic in general and for Web based services in particular. The QoE for a user of web-based services not only depends on network features but also on higher layer characteristics. Based on this observation, a method for objectively assessing QoE is needed.
Since QoE relates to a users’ experience, it partly involves a form of psychological measurement (subjective); however, it is important to telecommunication service providers to express QoE parameters in relation to network performance measurements
and equipment, which are objective in nature. There have been a number of investigations into this kind of relationship that are reported in the literature, [24, 27, 51] relating objective networking service conditions and user perception. This is an obvious way to provide a measure of QoE for users, and a suitable approach for establishing the relationship between QoE and QoS. In addition, Pais, I. [102] confirmed that QoS, QoE, user actions, and end user satisfaction can be established as a closed relationship loop as shown in Figure 4-1. From this figure, it can be seen that network performance, end user action, end user experience and end user satisfactior are closely related in which user experience is linked to end user action via network performance, and users perception is contributed to further network performance. This loop conveys the idea of the recent concept of quality that should be specifically based on the end user. Pais, I. research conveys the idea that the end user experience is not only related to network performance, but it also reflects their expectations and actions at the time of accessing the system. This shows that an end user’s experience is complicated, and needs more integrated information to understand, rather than a single parameter such as a network performance measurement or subjective user statisfaction scores. Basing QoE on only a single parameter could easily lead to a biased view of a user’s quality of experience or be difficult to assess their scores in more complicated service performance scenarios.
Figure 4-1: A closed relationship between QoS, QoE, end user action, and user satisfaction
Therefore, in a real integrated network environment, it is difficult to represent features of the various services using only the bandwidth and latency time; however, based on the higher layer characteristics of multi-media services, an analysis of service platform performance can be carried out. Any new model should have application-network- related performance criteria that correlate well with known or exact MOS scores. As
End User Action End User Experience End User Satisfaction QoS: Network Performance
factors that could affect outcomes of observations in order to avoid the pitfalls observed for an MOS that were based on scores and user surveys. As mentioned in the review discussed in Section 2.1, subjective assessments are prone to various errors because users’ scores are subjective in nature. Therefore, parameters of any new model should be based mainly on objective measurements, but they should avoid the complexities associated with, for example, the E-model that is based on impairment values [21] . The new model also needs to take into account any human factors, to ensure that the model can be aligned with MOS or similar subjective measurements of service quality.
4.1.2 Design of a suitable QoE model
From the comprehensive review presented in Section 2.4, there are typically three main approaches for measuring QoE. They can be divided into the following categories: firstly, a subjective quality evaluation [12] which is based upon statistical methods, user scores and user surveys; secondly, an objective quality evaluation [21, 22, 46]; and third, correlation between QoS and QoE [49-51].
Our novel model combines these three main approaches as mentioned in Section 1.4 in order to achieve two main goals that are required in the existing literature for a new model:
The QoE model should correlate well (better than 92.1%) with the Mean Opinion Score approach. The best correlation that has been achieved for QoE of web based services 2
(R ) is 0.921 based on session time [27].
Observations should be taken in an objective manner to infer the QoE being experienced by users [11, 51].
To build a model that satisfies our requirements, existing methodologies and models for QoE have been analysed. As previously mentioned, the above approaches all have their own drawbacks and merits as discussed in more detail as part of Chapter 2. Our novel approach that combines all three methods draws upon the merits of both of objective and subjective methods, and includes an understanding of the psychological aspects of users in order as described in Section 1.4, to produce some important enhancements. Firstly, using a QoS-QoE correlation approach to directly assess QoE based on current QoS information flows that have been recently captured and analysed; It can be said
that, inputs for the QoS-QoE correlation are real (objective) and based on current networking conditions and not predicted from only subjective inputs. Secondly, that QoS-QoE is accessed concurrently with the information reflected from the application layer in which the web service can be executed. Last but not least, human factors and content/time (CT) effects are taken into account in the model to provide a deeper understanding of a users’ perception.