End-To-End Quality Of User Experiment For Streaming Video Services Over Mobile Network

As the increasing bandwidth of mobile network and rapid development of streaming video technology, different kinds of streaming video services, such as video conferencing and video-on-demand, have been more and more popular. Because of the best effort service of IP network and unreliable transmission of UDP, current streaming video services cannot satisfy the requirements of users. In order to measure the user acceptance of a service, the industry considers Quality of user experience (QoE) as an important index. Real-time monitoring of QoE can help the network provider discover the network congestion and adjust the network resources in time. At the same time, service provider can improve their service level, with the assistance of QoE. Therefore, an end-to-end QoE assessment model for streaming videos over mobile network is needed.Firstly, this paper compares the advantages and application scenarios of the subjective and objective assessment method. The subjective experiment needs the participation of people and time-consuming, which makes it unsuitable for the real-time monitoring system. But the result of subjective method is with the merit of accurate, which can be used to train and verify the objective assessment models. Objective assessment method assess the QoE by building mathematical models with the used of distortion indexes. However, the number of distortion index in traditional objective assessment methods is restricted by the form of the mathematical formula and fitting algorithm. Therefore, the statistical learning method in the proposed QoE assessment model.Based on the above study, this paper defines25feature parameters including video-related features, end-to-end packetloss network distortion and end-to-end delay network distortion. The video-related features consider both the encoding distortion and video content. The end-to-end packetloss and delay network distortion features are extracted from both the packet level in the network layer and video frame level in the application layer.Then, this paper builds a mobile network end-to-end streaming video service simulation platform and performs a subjective experiment. This platform simulates the end-to-end transmission of streaming videos over real network and provides test videos with different degree of packetloss and delay network distortions for the subjective experiment. The result of the subjective experiment can be used in the training and verification of the QoE assessment model.Finally, this paper proposes a QoE assessment model based on decision tree method with the above feature parameters...The proposed QoE assessment model with the advantages of low computational complexity can meet the requirement of real-time monitoring. By comparing the resulting decision tree model with different feature parameter subsets, the impact of feature parameters can be well studied.