The Research And Application Of Real-Time Video Quality Assessment Based On Network Injured And Video Character Using BP Neural Network

As one of the landmark apphcatlons of triple play,IPIV is developing rapidly in the worldwide.IPTV has the character of large bandwidth and real-time,which require high quality for IP network.However,IPTV runs in the "best effort" IP network,which makes the users' quality of experience(QoE) be not guaranteed because of congestion,and result in losing potential users.On the process of replacing the traditional TV to become a mainstream,users want to access the video quality and user experience as good as traditional TV or even better.At the same time,IPTV operators need a real-time video quality assessment method to evaluate the level of their service.For the existing video quality assessment methods,subjective assessment methods are accuracy but wasting much human and material resource while objective assessment methods are simple but cannot reflect the users' true experiences.Against the existing video quality assessment methods' deficiencies,this paper presents a video quality assessment method(video quality assessment based on Network Injured and Video Character using BP neural network).It is as accurate as subjective assessment methods and as automatic as objective assessment methods,combined with BP neural network method to evaluate the video quality.It is a no-reference video quality assessment method,which analyzes the compressed domain of received video,extracts time and space characters as well as frame rate for quality assessment in the sample time.It is suitable for real-time evaluation of video quality,which can be used in quality control in IPTV video services.The characteristics of this approach are as follows:1.The subjective and objective video quality evaluations are of fully integrated, which both have the accuracy of subjective evaluation and the automaticity features of objective assessment.2.Extract the compressed domain of video sequence to analyze,without reference to the original video.It is a no-reference video quality assessment method, which can be used in real-time IPTV video quality assessment.3.With the BP neural network method to establish the relationship between parameters and the video quality,the accuracy is good,and the feasibility of this method is high,without complex operations.The main results of this study can be summarized as follows:1.Present a video quality evaluation model based on network injury and video characters,and give the realization of this model,using simulation methods to establish the relationship between QoE and parameters.2.Through the method of simulation,it creates the relationship between Video Quality of Experience and various parameters(video characters,frame rate).3.Extract two video characters:time-variable and space-complexity of their own to analysis,and gives the calculation of time-variable and space-complexity in the sampling time.4.Give the calculation method of actual video frame rate during the decoding (i.e.,the normal frames decoded in the sampling time).5.Based on the above work,an IPTV video service quality monitoring system is proposed which realizes the function by using the method described in this paper.This paper is supported by National Science & Technology Ministry of China (NO.2007BAH09B04) and Significant Scientific & Technological Project of Shanghai Science & Technology Commission(NO.08DZ 15001[10]).