Research On QoE Management Of Video Streaming Services For UDP Networks

With the popularity of video streaming media business and high-resolution video,users' demand on fluency and resolution of video have gradually risen,and therefore,operators and providers have come to attach more importance to the quality of video services.Meanwhile,the traditional QoS based business quality evaluation system cannot reflect users' actual viewing satisfaction properly.Therefore,the user experience quality QoE(Quality of Experience)for video streaming media business has become the research hotspot of relevant fields.Compared with traditional business quality management methods that are based on objective quality indicators,QoE pay more attention to the subjective and real experience of the user.In the process of evaluation,users mark their satisfaction on actual experience quality,and then we obtain the QoE assessment outcome.In actual business,however,the direct evaluation of users' subjective experience costs a lot.Therefore,the use of objective quality indicators is the main method currently used in practical application.As the result,it is the research priority for current time that using the measurement on objective quality indicators to sense user's subjective experience quality is able to overcome the defect of subjective assessment methods.In this paper,we study the objective QoE evaluation method of video streaming business,and propose an evaluation method for standard definition video business.Based on this,we also propose the evaluation,prediction and optimization methods for high definition video services.By establishing a hierarchical quality index system that reflects the end-to-end characteristics of video services,it not only evaluates the quality of video playing,but also reflects the influence of users' expectation on subjective experience quality in the assessment outcome.Specific studies are as follows:1)For standard definition video services,an objective QoE assessment method based on compression coding impairment and network transmission impairment is proposed.First,in the experimental environment,subject/objective assessment experiments are performed on sample video streams with different objective quality indicators(compression coding noise,bit rate,network packet loss rate,and temporal perceptual information).Through the comparative analysis of the evaluation results,it is verified that the image noise generated in the compression coding process will affect the user's subjective expectations,and then affect the final QoE evaluation results.Then use the experimental results to train and establish a video QoE assessment neural network.Under the condition of known objective quality indicators,the QoE quality of the current video stream can be evaluated.It is verified that the Spearman rank correlation coefficient(SROCC)between this assessment method and actual user experience quality is 0.9531 for the experimental environment of this study.For the public video QoE database EPFL-PoliMI,the Pearson linear correlation coefficient(PLCC)between the present evaluation method and the actual user experience quality is 0.9402,which is superior to other no-reference methods using this open database.2)For high-definition video streaming media services,a QoE assessment method based on image cumulative damage time is proposed.In the process of video playback,the cumulative time length of image damage events is used as an objective index for evaluating QoE quality.This method solves the problem that the perception of degraded image quality does not conform to the user's actual experience during the objective evaluation process based on the degree of damage of the frame image.In the study,we first discussed the reasons of cumulative damage time as an objective QoE evaluation index in HD video scenes.Further,the influence of image cumulative damage time on user experience quality is analyzed and verified by subjective/objective evaluation in the experimental environment.Then,the experimental data were used to fit the mapping relationship between image cumulative damage time and MOS score under different CRF parameters.Finally,using the fitting function to verify the QoE evaluation results,the verification results show that under different CRF parameters,the PLCC and SROCC coefficients between the evaluation results of this paper and the actual user ratings are better than the full reference assessment method SSIM results under the same conditions.3)For high definition video streaming services,on the basis of mapping relationship between cumulative damage time and MOS score,a QoE quality prediction model based on Bayesian network is proposed.First,using the end-to-end characteristics of video streaming services,we build a Bayesian network that can reflect the QoE quality of video services.In the Bayesian network,the network nodes represent the objective/subjective quality indicators that affect the quality of video QoE in theprocess of HD video streaming(Frame data quantity,network packet loss rate,frame data damage rate,image cumulative damage time,subjective MOS score)and the connection between nodes can reflect the relationship between these quality indicators.In Bayesian networks,the state of the child nodes can be predicted under the condition of known parent nodes.That is,under the condition of known the objective performance index,the cumulative damage time length of the image generated during the broadcast process can be predicted,and further user experience MOS score can also be predicted.In the experimental environment,the cumulative damage time of the image in the actual sample accords with the prediction result of the Bayesian network.At the same time,there is a good correlation between the sample subjective score MOS result and the prediction result.4)On the basis of established evaluation methods and prediction models,we propose an optimization strategy based on compression coding parameter control for high definition video streaming services.In the compression coding process,we generate video stream data adapting to the current network transmission state by controlling the bit rate and frame sequence structure of the video stream.Under the premise of controlling the intensity of image noise,we can reduce the probability of image damage,and optimize the quality of QoE.In the study,we first compare the data volume of sample video stream and image noise generated under different bit rate and frame sequence structure conditions.At the same time,we measured the cumulative damage of the sample video stream and the subjective experience quality of the sample video stream after the simulation network transmission through the network that has controllable packet loss rate.Further,by comparing the objective measurement and the subjective score,the effect of the bit rate and frame sequence structure control on the user experience quality is verified.Under the experimental environment,by controlling the compression coding parameters,the cumulative damage time of the image in the sample video stream can be effectively reduced,and the user's subjective quality of experience is improved.