Bayesian structural equation modeling of multidimensional QoE in haptic-audiovisual interactive communications
Shuji Tasaka · 2016
This paper proposes a methodology for assessing multidimensional QoE (Quality of Experience) by Bayesian analysis of Structural Equation Models (SEMs). As an example of the assessment, we employ a haptic-audiovisual interactive communication system because of its QoE's multidimensionality due to different traffic characteristics between the haptic media and audio-video streams. We deal with 14 dimensional QoE measures which consist of 13 subjective measures on a five-point scale and a single objective measure of efficiency; they are correlated with each other. In order to simplify the multidimensional causal relationships, we classify the QoE measures into three groups corresponding to three constructs (latent variables): AVQ (AudioVisual Quality), HQ (Haptic Quality) and UXQ (User experience Quality). We build a Bayesian SEM with the three latent variables and the 14 observed variables (QoE measures). We infer posterior probabilities of parameters in the model by Markov chain Monte Carlo (MCMC) simulation. As a result, we find that UXQ, which includes the QoE measure of overall satisfaction as an observed variable, is dominated by HQ and that the effect of AVQ on UXQ is small. We also notice that although the objective QoE (efficiency) is correlated with subjected QoE, a separate measure is necessary.