QoE Estimation Model For Video Streaming Service And Its Application In Wireless Transmission Control

With the evolution of mobile communication techonologies and the upgrading of smart terminal, wireless HTTP video streaming service is experiencing an exponential growth and is expect to be the new profit point for telecom operators and content providers. As users’satisfaction is the most important indicator for quality measurement, how to evaluate the users’Quality of Experience (QoE) about the service and adjust the policies of transmission control is becoming an important topic for industry and academia. In this paper, we investigate the QoE estimation model and its application in wireless transmission control.To the task of QoE estimation model, we group the comprehensive QoE influence factors into two types, namely the objectivity-aware parameters and psychology-aware parameters. The considered factors include video content features, the encoding parameters, the network transmission metrics and the playout buffer parameters. Moreover, we use a multiplicative fusion method to combine the impact of different parameters on QoE, and establish a QoE estimation model for HTTP video streaming service based on support vector machine (SVM). Simulation results show that the proposed SVM-QoE model performs well in terms of high Pearson correlation coefficient and low root mean square error. Meanwhile, the SVM-QoE model has a low computational complexity that can be implemented in realtime estimation system.To the task of QoE-based transmission control, we choose scalable video coding as the source enoding method and take the long-term expected QoE as the optimization objective. Then, we formulate the problem of wireless video transmission by a partially observable markov decision process (POMDP) framework, and proposed a QoE-based rate adaptation scheme for video streaming service. Meanwhile, an online learning algorithm is proposed to solve the POMDP on the fly. Simulation results show that through adjusting the video rate periodically, the proposed QoE-Learning rate adaptation scheme can effectively release the network congestion, improve the users’ QoE and video smoothness. Meanwhile, it performs well under different bandwidths.Using the SVM-QoE model and QoE-Learning rate adaptation scheme, telecom operators and content providers can estimate the users’ QoE effectively with the given informations, and adjust the video rate appropriately to improve the users’ satisfaction, which is meaningful for commercial applications.