Research On Video QoE Evaluation Based On Mobile Crowd Sensing

The mobile streaming media service has been explosive growth in recent years,and has become one of the most popular services in mobile Internet business.With the huge market,there is fierce competition between operators and streaming media service providers.In the user centric business era,researching on how to improve the quality of service for users,ensuring that the service can be recognized by the user will play a decisive role in the competition.With its vigorous development,video streaming quality of user experience evaluation still faces two major defects:low efficiency,time-consuming data acquisition;poor real-time and low integrity evaluation result presentation.Therefore,it is necessary to improve the way of data collection,and establish a complete quality platform for user experience,which can display the data in real time.This paper introduces the Wilysis system based on mobile crowd sensing,and attempts to improve the QoE evaluation model to increase the robustness and accuracy of QoE prediction.The main contents of this paper are as follows:Firstly,this paper surveys the current situation of mobile video services,development trend of video QoE and mobile crowd sensing,analyzes the current research of video QoE may face the problem:the cost of data acquisition is high,the imperfection of the collection platform leads to the failure of data to form a closed loop and the real time QoE can not be guaranteed.In this paper,the definition,quantification method and rating method of video QoE are briefly introduced and the basic characteristics,application fields and general structure of mobile crowd sensing are introduced.Finally,the main innovation points of this paper are summarizedSecondly,based on the the concept of mobile crowd sensing,this paper presents the Wilysis computational model,including crowd sensing model,QoE evaluation model.Based on this computing model,design and implementation of the Wilysis system based on mobile crowd sensing,describes the overall architecture of the system and the calculation process of the QoE evaluation model in the system.The feasibility of the model and the system is proved by designing the test schemeFinally,the paper analyzes the applicable scenarios of the QoE evaluation model,and summarizes its shortcomings in the environment of massive data:linear,non data,and unable to modify the influence factors and its weight coefficient.According to the data collected in the Wilysis system,and in view of the above three points,try to optimize the QoE algorithm,we design the polynomial regression,QoE-elastic BP neural network,QoE-tree model fusion three schemes,and describes the establishment process of the three schemes briefly.Through experiments,the advantages and disadvantages of various indexes of the original model and the improved model are verified and discussed.Finally,the best fusion scheme of QoE-tree model is obtained.