Quality of experience for adaptive streaming using HTTP Dynamic Streaming

Abstract: This thesis investigates how to establish the relationship between OSI layer 7 parameters of video streaming and the QoE of the user, and to evaluate which methods are most tting for the estimation of QoE. The project is made in cooperation with LTH and Acreo, and is a part of the Next generation over-the-top multimedia services (NOTTS) 7and the Eco system for Future Media Distribution (EFRAIM) project. The underlying techniques, which form the environment of our research of estimating the QoE, is adaptive bitrate streaming over TCP. The purpose is to investigate how a service, that provides a user with the means to benchmark the received quality of the Over the top (OTT) streaming service, can be built and distributed. Today there exists no such service that takes the viewers subjective opinion into consideration. There have been extensive research on some connected elds and issues but none with a unied solution to streaming adaptive bitrate video over TCP with its particular behavior and eect caused on the streamed video. In this report we evaluated two dierent methods of prediction of QoE, Pause Intensity based on the number of pauses and their length during playback, and a Linear bitrate model based on the average bitrate quality and its standard deviation. We also made a small user test with our streaming client software to evaluate the two methods to decide which one is the most benecial to use. The test showed that although one of the most irritating playback deficiencies is when pauses occur, the linear bitrate model delivered the most accurate predictions.