Research And Application Of Quality Of Experience Measurement Method For Network Video

In the past few years,online video services have rapidly gained popularity.The increasement in access network speed and convenience are also some of the reasons why a large number of users choose to watch online video.The flow of online video streaming increased from 57%of total traffic in 2012 to 69%of total traffic in 2017.In order to retain existing users and attract more new users,video service providers try to satisfy the user's expectations and ensure provide a satisfying viewing experience for the user.The first step in providing satisfactory services is to be able to quantify how users feel about the current level of service.Quality of Experience is an indicator that can fully evaluate the sensory quality of users.This thesis has conducted research work on network video quality of experience from four aspects:(1)After our research and discussion,we established 14 video quality of experience parameters required for this study and established the method of collecting these parameters.These parameter indicators include:the total duration of the video?the duration of video playback?the number of dragging?the total duration of the drag?the drag frequency?the number of pauses?the total duration of pause?the pause frequency?the number of buffers?the total duration of buffer?the buffering rate?the start time of the video?video end time and user ratings.(2)For the 14 established experience metrics,we have developed a system software for measuring network video playback parameters using technologies such as Spring Framework,Spring MVC Framework,MyBatis Framework,and HTML5.The core work in this measurement system is to develop a web video player that can automatically measure 14 QoE parameters in the background through HTML5 technology,JavaScript technology,and jQuery technology.The system's measurement process is transparent to the user and the process does not affect the quality of network video playback.(3)In order to study the relationship between the 14 video quality of experience parameters that we have established,we have designed a new process of subjective and objective assessment.In this process,a multi-dimensional linear regression method was used to create a QoE evaluation model for network video QoE.The evaluation model reflects the linear relationship between multiple objective QoE parameters and user experience quality.(4)In order to help video service providers understand the user's video viewing behavior.We use the gradient-elevation decision tree method in machine learning methods to create a predictive model of user viewing behavior.The predictive model can predict the behavior of the user's rating behavior,closing the video behavior and dragging the video behavior.After a lot of experimental results,the accuracy rate of our proposed network video experience quality assessment model reached 79.9%,and the accuracy of our proposed user viewing behavior prediction model reached 94.2%.The evaluation model and prediction model proposed in this paper are closer to the real feedback of the video users.The research in this article helps network video service providers to improve their service quality to attract more users.