Research Of Multi-Features-based Video Quality (QoE) Metric Under Wireless Network

With the rapid development of mobile communication technology, the wireless bandwidth is growing, which makes high-speed video services account for most of the wireless communications traffic flow. However, due to the diverse types of video services, simple relationship does not exist between the traditional network performance evaluation and video quality. In addition, the video quality descending is much more complex than voice services; QoS parameters used to describe the voice quality cannot describe the video quality accurately. So a video quality evaluation system is required by inducing the video characteristics, which can cover the whole wireless network and meet the different needs. What is more, utilization rate of wireless network resources and the video quality will be improved by using this system model.The above problems are investigated in depth in this paper with the support of the National Natural Science Foundation Program of China and Beijing Science And Technology Plan Program. The main innovative contributions are listed as follows.(1) Based on the study of sequence-based pixel-level feature extraction, a no-reference video quality assessment model is introduced. This model uses the concept of cross-layer design, fully integrated video features characteristics of the application layer and network layer characteristics. This model has four main contributions:1) we induce the impacts of content-aware features to the video quality estimation to ensure the video quality estimation accuacy;2) PCA is used to network layer parameters to reduce the network layer parameters which reduce the video quality computational complexity;3) SVM is chosen to merge all the cross layer parameters to predict the perception scores of video quality which impove the caclution polymerization;4) multiple video features, application layer parameters and network layer parameters are normalized to solve the SVM’s paranoia which improve the accuracy of the video assessment.(2) Based on human eye simulation studies, two video quality prediction metrics are provided which is suitable for heterogeneous wireless networks. Meanwhile, these metrics expand the video quality evaluation scope and improve the video quality prediction accuracy. These models have three main contributions:1) On the basis of physiological characteristics and psychological characteristics of the human visual system, DCT converted JND is used to descripe the video feature which improves the video quality prediction acccuray under heterogeneous network;2) both JND and QoS are used to reduce computational complexity of new parameters extraction;3) QoE prediction model is divided into two scopes model as special model and cross-layer model, which can adjust the network performance or/and encoding parameters.(3) Based on the video coding techniques, a no-reference video quality evaluation model is proposed under wireless network which is good at real-time, accuracy and complexity. This model has two main contributions:1) the parameters used in the model (such as coding and transmission factors) are adopted directly from the video stream, which reduce the decoding process;2) by analyzing the characteristics of intra-coded frames and inter-frame coded frames, we extract factors related to video quality decline both from the temporal and spatial which improves the accuracy of the dynamic video quality assessment(4) On the basis of characteristics of video coding and the wireless/mobile network functional entities, a no-reference video quality prediction model is provided for mobile terminals, the main contribution can be included as following two aspects:1) analyzed the relationship between wireless network parameters and the video quality, a lightweight classification video quality prediction model is given based on QoS parameters;2) a simple video classification is provided from the intra-coded frames from video bitstream, and has be proved to be having strong discrimination and robustness. In the performance analysis, the model has good stability, accuracy and real-time.To sum up, the entire model system covering the various stages of the wireless video transmission and wireless access with good performance, and achieve the expected objectives of this study. The dissertation have some breakthrough in theory, some innovation in technology and some reference in methodology, it opens up a new way for wireless video quality assessment.