Research Of Video QOE Based On Objective And Subjective Factors

With the popularity of smart devices and striding development of mobile Internet, public life has been reshaped. Large number of applications emerge and take up much of human time. However, time of user is limited, which leads to intense competitions of service providers.Under the condition of limited resource and homogeneity service, good Quality of Experience (QoE) is the key to to win more users in the competition. Of all the research directions, QoE evaluation is the most basic and important. Meanwhile, QoE evaluation has a strong reference value for business adjustment in the actual scenarios. Many experts and scholars have conducted vast researches on QoE evaluation and established some excellent QoE models. However, the factors almost focus on a single point, which does not reflect properties of QoE comprehensively. In view of this, this thesis evaluates QoE synthetically from the point of subjective factors and objective factors. Main contents are as follows.First, this thesis analyses and summarizes current QoE evaluation methods, influence factors and common QoE evaluation models. Then,from the perspective of the research status and future development trend of application, we chosen the hybrid evaluation method and settled the scope of influence factors.Second, we analyses the influence of factors respectively from the point of subjective psychology and objective service quality. Based on the analysis, we divide influence factors into two parts: subjective factors and objective factors. The subjective factor is mainly composed of user preference. Objective factors include video image quality, network parameters and buffer information. Based on influence factors, this thesis designs and builds the experiment platform which is composed of terminal application and servers. Then, we implement the experiment of collecting QoE data which will be applied to the QoE model based on the platform.Finally we analyzes the corresponding relations between factors and QoE. According to the result of data analysis, we divide the model of QoE into two parts: subjective model and objective model. For subjective model,its output is the QoE value in ideal conditions, which represents users'expectation to some extent. The result of objective model is the quality rating of objective service. The final QoE value is computed by combining the results of subjective model and objective model together and applying a joint decision function. When solving the subjective model, the method of linear regression is used. The Softmax model of multivariate classification and BFGS method is applied to the objective model. Finally QoE data collected by the platform is applied to the proposed model.Results show that more than 90% of the QoE is predicted effectively, which verifies the validity of the proposed model.