Research On QoE Of MPEG-DASH Streaming Media Based On Android Platform

With the popularization of Internet and the great enrichment of online entertainment contents,people are more and more inclined to watch video online.However,traditional streaming media technologies have various disadvantages,RTSP/RTP streaming technology for example,it requires specific streaming server,which are complex on both installation and maintenance;furthermore,it cannot penetrate firewalls.Another example is the HTTP progressive downloads,it does not have the bitrate selection options,it can either cause a pause or waste of bandwidth if the network speed fluctuates.MPEG-DASH(Dynamic Adaptive Streaming over HTTP),combines the advantages of HTTP progressive download technology and Adaptive transmission technology,can well solve these problems.This thesis mainly studied the adaptive algorithm of MPEG-DASH streaming media network with Android OS.Based on embedded media box environment,aimed on the improvement of user Experience QoE(Quality of Experience),this thesis studied in detail on below items,startup delay,buffer underflow,average code rate and total code rate switching times,etc;and proposed optimized scheme accordingly.Firstly,this thesis introduced the history of MPEG-DASH and the current research status of adaptive algorithms.Then,it described in details of the technical standards for MPEG-DASH adaptive streaming media,including the technical system framework of MPEG-DASH,the composition,structure and the examples of MPD files,the adaptive transmission process of MPEG-DASH and its shortcomings.Then,based on the items' optimization and improvement mentioned above,which include the startup delay,buffer underflow times,average code rate,total code rate switching times and other aspects,we mainly did the following work:1)In terms of shortening startup delay,we optimized the FFMPEG DASH Encoder/GPAC MP4 BOX DASH Encoder,added the SPS/PPS information,known as Codec Specific Data(CSD),which is required for decoder initialization,to the media presentation description document(MPD)as a subfield.We can then parse this information in advance on the client side and initialize the decoder in parallel;thus,we get progress on improving the startup speed.2)Considering that the DASH usually adopts the H264 video encoding based on dynamic bitrate coding,in some cases when the video sources have the high bitrate volatility,mis-judgement may occur when judging whether the bandwidth is sufficient.In this thesis,we took use of the sidx box item of the initial segment,which stored whole video segments' information,to calculate the average segment bitrate of all segments.When playing back the video,the bitrate of the next segment is selected based on the bitrate of the next several segments to be downloaded combined with the current network bandwidth and the length of buffers.As the average segment bitrate is sometimes more than 100% higher than the average whole video source bitrate,it is more practical to take into account of the segment bitrate at the same time during the real using situation.3)Based on Android 9.0 client,ExoPlayer APK,this thesis implemented and verified the optimization algorithm: build a LAN Apache HTTP server on Ubuntu 18.04,add network bandwidth limitation to simulate different network scenes.We can find that with the optimized algorithm proposed in this thesis,we got significant improvement at the startup delay,the times of underflow for the playback buffer,the frequency of switching between different bitrate sources and other aspects;In terms of comprehensive QoE index,among the several algorithms,the NextN algorithm proposed in this paper has the highest QoE value in various bandwidth-limited scenarios of the two experimental video sources,which confirmed the effectiveness of the method in this thesis.