Evaluating Quality of Experience in Videoconferencing Systems Based on Packet Loss

Abstract

Users of multimedia applications get more and more quality aware as the development of technologies progresses. When looking at videoconferencing it would be a great advantage knowing what level of quality a user experiences at any given time. By having this ability, a service provider might be able to improve the quality for the user by for example increasing the bandwidth used for the call. Traditionally video quality has been calculated by comparing the sent and the received video streams. This thesis focuses on estimating experienced user quality based on metrics collected by a supervision system. These metrics might be packet loss, degree of compression, available bitrate and so on. A model is created based on statistics gathered from a set of data provided by Tandberg. It is shown that for high bitrates the quality can be estimated with acceptable precision, while for lower call rates like 128 kbps it is more challenging. A demonstration application is implemented to give an example of how the model may be utilized. At the end of the thesis, some proposals are given as to how one can improve on this model, increasing its precision and subsequently improving the metric denoting the experienced user quality.