| The new technology of mobile communication has provided nutritious soil for the growth of mobile services, but also invoked the user's expectation for higher quality service. To survive and prosper in the ever-changing service landscape, service providers seriously consider Quality of Experience (QoE). QoE concerns service quality from the user's perspective.;This dissertation reports our research to analyze and assess QoE of the Over-The-Top (OTT) video service. It has proposed to study QoE based on the life cycle of a video session. Based on the concept of a life cycle, it classifies interruptions during a video playback as impairments (temporary interruptions) and failures (permanent interruptions).;This dissertation introduces a detailed subjective methodology for assessing session-based QoE in a laboratory controllable approach. The methodology includes selection of video content, design and implement of test conditions, delivery of questionnaire, to data collection. It is eminently useful for service providers and researchers who are interested in assessing video QoE. Three laboratory controllable experiments have been conducted following this methodology.;This dissertation also provides QoE objective assessment based on data collected from these subjective experiments. First, the dissertation presents the impact of failures on the QoE evaluation by statistical tests, which explains why we need to study session-based QoE. Second, it discusses the interplay among selecting QoE factors, CQ (Content Quality), TQ (Technical Quality), OX (Overall eXperiment), and acceptability. It demonstrates that TQ is the determinant factor of OX with the presence of failures in these laboratory controllable experiments, which is valuable for future session-based QoE studies. And then, the dissertation investigates the relationship between QoE factors and Application Performance Metrics (APMs). Proposing novel APMs to predict QoE factors, the feasibility of APMs is examined and a primary QoE model is proposed. Lastly, this dissertation identifies that the user's rating behaviors have different levels of sensitivity to impairments and failures. Principle Component Analysis (PCA) is proposed to discern abnormal behaviors and classify user personality in QoE assessment. |
