QoE Analysis for Interactive Internet Applications in the Presence of Delay

Quality-of-experience (QoE) of interactive applications transmitted over TCP/IP networks has recently gained considerable attention, and is mainly influenced by transmission delays due to TCP/IP?s retransmission characteristic. This thesis shows that interactive Internet applications share the commonality of a recurring request- response cycle that is highly vulnerable to such transmission delays. In the context of two prototypical services, interactive Internet telephony and browser based applica- tions, the impact of transmission delays on QoE is analysed. In terms of interactive Internet telephony, a surface structure analysis of delay impaired voice calls reveals several changes in conversation behaviour caused by the delay. From this analy- sis, two conversational metrics are derived that capture the influence of delay on human-to-human conversations. Using these metrics as additional input parame- ters, an update to the E-Model is proposed that enhances prediction performance considerably. For browser based applications, a novel subjective testing method- ology is presented that establishes a realistic flow-experience in the resulting web browsing sessions. Data from two lab studies and a field trial proves the ability of this test methodology to provide reliable and consistent results across di?erent con- texts. In terms of the relationship between waiting time and QoE for browser based applications, this thesis postulates the WQL hypothesis: the relationship between ?Waiting time and resulting QoE is Logarithmic?. With the acquired data from the three studies, the WQL is verified for file downloads and simple web browsing. Contrary, in the context of complex web browsing the WQL has to be rejected. A following analysis reveals several challenges and practical issues that complicate the use of the WQL for this service. Additionally, it identifies the subjectively perceived page-load-time as an interaction based measure of waiting time and promising input parameter for novel QoE models. Finally, a human perception model, that considers interaction quality performance aspects in the quality formation process, and that explains (re-)actions to (conversational) input signals in the form of active output signals is derived.