End-to-end quality adaptation scheme based on QoE prediction for video streaming service in LTE networks

How to measure the user's feeling about mobile video service and to improve the quality of experience (QoE), has become a concern of network operators and service providers. In this paper, we first investigate the QoE evaluation method for video streaming over Long-Term Evolution (LTE) networks, and propose an end-to-end video quality prediction model based on the gradient boosting machine. In the proposed QoE prediction model, cross-layer parameters extracted from the network layer, the application layer, video content and user equipment are taken into account. Validation results show that our proposed model outperforms ITU-T G.1070 model with a smaller root mean squared error and a higher Pearson correlation coefficient. Second, a window-based bit rate adaptation scheme, which is implemented in the video streaming server, is proposed to improve the quality of video streaming service in LTE networks. In the proposed scheme, the encoding bit rate is adjusted according to two control parameters, the value of predicted QoE and the feedback congestion state of the network. Simulation results show that our proposed end-to-end quality adaptation scheme efficiently improves user-perceived quality compared to the scenarios with fixed bit rates.