| User’s quality of experience(Qo E) evaluation is one the most important issue in the research of video service quality guarantee. The information world of human living in is a huge and complicated communication ecosystem, where user’s experience will be interactively affected by various factors in different domains such as technology, business, context, and human. How to accurately analyze the relationship between various factors and Qo E is a difficulty in Qo E evaluation. In this dissertation, user’s Qo E evaluation method for the video service is explored. Targeting the different application scenarios, the networked video quality assessment is mainly studied from the perspective of the technology domain of communication ecosystem, and a parametric planning model, a packet layer assessment model and a bitstream layer model are proposed, respectively. In addition, the joint influence of user domain and technology domain on Qo E is also studied and an objective Qo E evaluation model is proposed. The major contributions of this dissertation are summarized as follows:1. To better plan and deploy the coding and network parameters of network video service, a parametric planning model for video quality assessment is proposed combing the characteristics of the channel and video sequence. The probability distribution of packet loss state is analyzed according to the characteristics of the four-state Markov channel. Then, combining the burst packet loss and temporal correlation between frames, the parameters that reflect the sequence distortion and frame distortion are derived. Using these parameters, the distortion of the video quality can be measured. Compared to the ITU-T standard G.1070 model and T-V model, the proposed model can estimate the distortion caused by directly packet loss and error propagation more accurately.2. To realize estimating the network video quality in real-time, a packet layer video quality assessment model is proposed considering the temporal characteristics of the video content. Through analysis of packet headers, the proposed model can obtain some basic information of the video and network. Then, a frame type detection method is proposed according to the coded bit-rate of each frame, where a dynamic threshold is employed. Considering the influence of motion activity of the video content on the video quality, the ratio of the average coded bit-rate of P frames and I frames is employed to estimate the temporal complexity of video content. Then, the coding distortion is evaluated combining the coded bit-rate and the temporal complexity, and the distortion caused by packet loss is further estimated combining the position of packet loss and the temporal complexity. Finally, the video quality is obtained. Experimental results show that the performance of the proposed model is better than that of ITU-T standard G.1070 model and DT model of Deutsche Telekom.3. To better estimate the influence of packet loss on the video quality, a bitstream layer video quality assessment model is proposed based on the frame quality. According to the different influence of packet loss on the frame quality, the frames are divided into four categories: frames without any distortion caused by packet loss, frames affected only by the loss of their reference frame, frames affected by error propagation without reference frame loss, and frames affected by both loss of their reference and error propagation. In addition, considering the characteristic of video content to the video quality, the temporal complexity of video content is estimated by differentiating the moving portion and statistic portion of the video content. Then, the quality of each frame is evaluated combining the temporal complexity of video content. Finally, the video quality can be obtained by a two-level temporal pooling scheme. Compared with another two models that based on frame quality, the proposed model has superior performance.4. To evaluate user’s experience on the video service more comprehensively, an objective Qo E evaluation method is proposed combining the user’s perceptions of video quality and video content. Different from the traditional Qo E evaluation method, the proposed model focus on the user’s perception of video quality and user’s interest in video content, and estimate user’s Qo E by combining these two basic aspects. To describe the user’s interest in the video content, a single camera-based viewing behavior detection method is proposed, which is synchronized with the watching process. Some unconscious viewing behaviors such as blink, fixation are analyzed and finally the blink is employed to estimate the user’s interest. Experimental results show that the ratio of the sequences whose predicted Qo E values are equal to the actual Qo E values is 64.5%, while the sequences whose predicted Qo E values are ±1 deviation with actual Qo E values is 98.8%.
|
PDF Full Text Request