Research On Video Measurement And Content Aware Network Bandwidth Allocation Mechanism In Wireless Network

Video service has account for amount of Internet traffic.According to CISCO's research report,video traffic of mobile terminal will occupy 82%of the total network traffic by the year 2020.Meanwhile,the number of video users has reached up to 545 million by report of Internet Development Report.Furthermore,quality of experience(QoE)is an important performance metric to measure users' satisfaction of video service.Therefore,with regard to networks theory and networks operation,it is significant to measure video's QoE in actual networks and improve users' QoE.The video experience of users is not only related to network transport parameters and video encoding parameters,such as bandwidth,delay,packet loss probability,frame rate and frame size,but also related to content type of video.With respect to the first field,this dissertation measures and analyzes the impact of device hardware,network transmission parameters and video encoding parameters on users experience based on large scale log data of video application.With respect to the second field,this dissertation proposes two optimal models to realize the content aware video rate allocation and one method which enable video streaming to share network bandwidth based on content type of video segments.The main contents of this dissertation are as follows:(1)We do statistical analysis and data mining based on large scale network video data.First of all,we do statistical analysis of large scale video data,including video length,start time of watching video,device type and device price.Secondly,we analyze these factors' impact on video completion ratio of users.Finally,we explore the relationship between these factor and video completion ratio based on k-means.(2)Two optimal models are proposed to realize the content aware video rate allocation.Previous research has shown that the video users experience is different under a same transmission condition such as bandwidth and loss rate if they watch different content types of video.For example,the users' experience of an action video watcher is lower than that of a romance video watcher.Therefore,network bandwidth allocation should take content type of video into account.Based on this observation,we propose two optimal models to realize the content aware video rate allocation.The numerical results of this dissertation show that the first model realizes fair sharing bandwidth in terms of users' quality of experience;and compared with traditional bandwidth allocation mechanism,second model increases the total MOS value of all users by 2.7%.(3)A network bandwidth sharing method is proposed based on content type of video segments.In above work,the classification of video type is categorized according to the entire video file.Actually,content type of video is dynamically changing with video playing.For example,fighting scenes and lyric scenes appear alternatively in one video.Based on this observation,this dissertation proposes a network bandwidth sharing method based on content type of video segments.The basic idea is as following:the video is sliced into segments,and network bandwidth is allocated according to content type of video segments.The numerical results show:in the movie samples adopted by this dissertation,in best situation,this method further increases the quality of experience by 0.25%at least and up to 0.92%compared with network bandwidth allocation method in second work.