Research On Modeling And Prediction Of Wireless Network Video Quality Of Experience (QoE)

With the rapid development of wireless network communication technology,mobile online video service has become the mainstream application of mobile Internet,and occupies most of the wireless communication traffic.At the same time,the user's experience demand for online video service under wireless network has changed gradually.Single subjective score and traditional QoE modeling scheme can not fully reflect the user's actual experience quality.Therefore,building a wireless network video experience quality prediction model that meets the current user needs has become a challenging hot issue.This paper relies on the National Natural Science Foundation of China's project "Research on the Measurement,Modeling and Optimizing Method of Network Video Experience Quality 61672221",to study the modeling and prediction methods of video experience quality under wireless network.The specific work and contributions are as follows:(1)Based on the analysis of the influencing parameters of video experience quality in wireless networks,it is clear that the influencing parameters can be divided into four categories: video information,quality of service(QoS),user behavior and facial expression.The video information established in this paper includes video length,frame width,frame height,frame rate,video type and video playback progress.QoS includes buffer times and total buffer time.User behavior includes the number of drag progress bars,drag progress bars,pause times,total pause time and user viewing time.Facial expressions are facial expression pictures captured when users watch video.(2)Design and implement a wireless network video QoE data acquisition system.The system consists of QoePlayer client and server application.The QoePlayer client can be deployed on Android mobile terminal,providing online video playback service and time-sharing QoE data acquisition.The server mainly has the functions of data reception and data storage.(3)Construct a sequence prediction model from multiple time steps of video session to QoE.Firstly,a QoE prediction model based on video information,QoS and user behavior is constructed.Based on this model,facial expression data are introduced when users watch video,and a QoE prediction model based on video information,QoS,user behavior and facial expression is constructed.Both models are implemented by multi-task.Prediction of multiple QoE indicators.By comparing the performance of the proposed QoE prediction model with that of the traditional QoE prediction model using the mean square error as the performance index,the experimental results show that the proposed QoE prediction model has smaller mean square error value and better effect than the traditional QoE prediction model,which is helpful to provide a more comprehensive and accurate QoE reference for video providers.