Video features with impact on user quality of experience

Abstract

One of the major challenges in wireless networks is to quantify and evaluate the user quality of experience (QoE). Hence, network operators are continuously working towards identifying the factors that most affect the user quality of experience for better network resource management. This paper proposes an evaluation method for video features with impact on user quality of experience. The goal is to extract and sort the features that highly affect the user experience and satisfaction while watching a video. The paper contributes first in determining the video parameters as well as the network features affecting QoE. Second, a survey is conducted to obtain subjective assessment on user satisfaction where users give their experience score and comments while watching videos with different feature values. We conduct experimental measurements and record videos with different frame rate, video resolution, transmission data rate, packet loss, delay and codec types. After collecting the QoE assessments, the supervised training data set is developed and imported into Rapid-Miner data mining tool. The feature evaluation is performed first using forward elimination feature selection algorithm, and second using decision tree classification to extract and sort the features that highly affect the subjective QoE. The results showed that the video resolution had the highest impact on user QoE. In addition, the transmission data rate was correlated to packet loss and delay and had higher impact than video frame rate and codec type.