Research On Evaluation And Analysis Of Network Video Quality Of Experience

With the continuous development of mobile communication capabilities and continuous improvement in the storage capacity of terminal devices,more and more people are opting to watch online TV via mobile devices.Network providers and service providers must ensure that end users receive their approval,namely good Quality of Experience(QoE).Therefore,the evaluation of user experience quality is not only one of the significant research topics in the academic community,but also the key to improve the service quality of network providers and service providers.The quality of user experience is a subjective experience of users watching videos.Relying on subjective tests to obtain QoE requires more manpower and material resources,and the test results obtained are posterior;while the objective test results of QoE are a priori,but the accuracy of its predictions is not guaranteed.This article focuses on the end-user video experience quality as the research object,and uses a combination of subjective and objective evaluation methods to study the user experience quality of video.First,we need the user's subjective quality of experience for the video service,and obtain some objective parameters in the user's video watching process.By analyzing the relationship between the objective parameters and the subjective experience quality,finally we establish a variety of assessment models which can accurately reflect user's video experience.The main work and innovation of this article are as follows:1.Subjective test data is inevitably subject to gross errors.Therefore,M-MCD is used to judge the gross error of QoE subjective test data.When the gross error is discriminated by Mahalanobis distance,the data difference in the data set is mixed in advance and the mean value and the covariance are distorted,which leads to the inaccuracy of the gross error discrimination.The MCD gross difference method uses the sample mean and covariance instead of the overall mean value and the covariance to determine the gross difference.However,there is a problem that the number of samples selected by MCD is difficult to determine.Therefore,the number of samples to be selected is determined using M-MCD first,and then the gross errors are determined using MCD.2.This paper proposes to use multiple nonlinear regression models to evaluate the user's video experience quality.We determined the data set's gross error,the average,standard deviation,correlation calculation and distribution function,finding that the data set has an approximate linear relationship.Therefore,we first established a multiple linear regression evaluation model.However,the linear model can not accurately reflect the quality of user experience.Therefore,a multiple nonlinear regression model was used to reflect QoE and a more satisfactory estimation result was obtained.3.The use of crowd search algorithm(SOA)to calculate the regression parameters of multiple nonlinear regression models.When the multiple nonlinear regression evaluation model is established,the calculation of regression parameters is more complex,involving more partial derivatives,matrix inversion,etc.The calculation is large and error-prone.Calculating the regression parameters by using the crowd search algorithm greatly reduces the programming complexity of the program,and the QoE accuracy estimated by combining SOA with multiple nonlinear regression is also satisfactory.4.A QoE evaluation method combining SOA and wavelet neural networks is proposed.Although the multiple nonlinear regression model estimation results are acceptable,this paper wants to further improve the estimation accuracy and improve the model performance.Because the wavelet neural network is also a nonlinear model,this paper proposes using the crowd search algorithm to optimize the wavelet neural network to subjective quality of user experience.Through the combination of SOA and wavelet neural network,the response QoE assessment model is finally obtained accurately.