Abstract :
This paper presents a Bayesian statistical approach to QoE (Quality of Experience) assessment problems in multimedia communications. In order to explain how to apply the Bayesian method to the problem, we pick up, as an example, an interactive audiovisual communication with a QoE enhancement mechanism, which we call user-assisted QoE enhancement. Utilizing experimental data measured on subjects (users) in the interactive communication system, we first set up a framework of the QoE assessment in the context of Bayesian modeling.We build a couple of Bayesian hierarchical models with two-levels, user and gender, based on multivariate ordinal categorical logistic regression using a latent variable. We then perform Markov Chain Monte Carlo (MCMC) simulations with WinBUGS and check the convergence to obtain posterior probabilities of model parameters. Utilizing the posterior probabilities, we calculate QoE measures with 95% credible intervals and compare them with the traditional MOS (Mean Opinion Score) with 95% confidence intervals. We also examine the effects of video bandwidth, the kinds of tasks, the user-assisted QoE enhancement mechanism on QoE by taking into consideration the subjects´ gender or equivalently his/her technical background of computer in this case.
