Research On Network Video Quality Assessment Method And Measure Technology

With the fast development of broadband communication technology, network video users grow rapidly. In order to obtain the recognition of network video users, video service providers need to provide high quality of user experience (QoE). Therefore, the network video quality assessment and measure technology have become an important research topic of the academia, but also the key technology of industry.Many research institutions and scholars have done some research on this field and achieved certain achievements. However there are many factors which impair the network video quality. The subjective network video quality assessment method has many complicated steps and long time, low real-time performance. The objective network video quality assessment method depends on measurement, and the similarity of objective and subjective should be improved. Network video quality assessment and measurement methods are still many problems to be solved.This paper has done some research on video quality assessment, and established the objective assessment model to improve the similarity between the objective and subjective assessment method. The research focus on analysis of network video quality under different experimental environment, extracting the network performance indexes, video characteristic parameters and network application indexes. This paper proposes the corresponding improvement objective assessment method, and verifies the method performance by theoretical analysis and measurement experiment. Meanwhile a set of network video quality measurement software is developed. The main work and innovations are as follows:1) A hybrid network video quality assessment method based on multivariate nonlinear regression is proposed. IPTV is selected as research object. Network bandwidth impairs video quality more when transmission link establishment. So the network video quality is measured under different network bandwidth. The hybrid assessment method fully considers the network performance, video parameters and combines the structural similarity (SSIM). It can get more accurate results than SSIM and improve the similarity between the subjective and objective assessment.2) An objective network video quality assessment method based on M5’model tree is proposed. In order to solve the media delivery index (MDI) only measures the network environment without objective result, the MDI is measured under different bandwidth and trained. Then the objective assessment method is proposed. This method can establish the mapping relationship between the MDI and QoE, and give the objective result.3) An objective network quality assessment method based on Mamdani inference is proposed. This method analyzes the relationship between the network layer QoS, application layer QoS and QoE. HTTP video is selected as research object. The three application indexes (initialization time, re-buffering time, re-buffering frequency) of HTTP video is measured under different network environment. According to the measured data, the data is arranged into different range. Then the rule table is designed to infer the fuzzy data. At last the objective score is got, using the Mamdani inference. This method can improve the similarity between the subjective and objective. On this base, how the TCP throughput impairs the network video quality is studied. In order to forecast the TCP throughput more accurate, an improved TCP throughput model is proposed. The experimental results under different network environment show that the improved model more close to the actual value.4) A comprehensive network video quality assessment model is proposed. This model considers different factors which impair the network video quality and the analytic hierarchy process (AHP) is used. According to the actual environment, this method can select different factors to establish the QoE model. It can improve the accuracy of the network video assessment. On the other hand, we comprehensively consider the method mentioned in this paper, and design a set of video quality measurement software. The measurement software can monitor the video indexes and give the objective results.