The Study Of Optimization Technology For MPEG-DASH Streaming

With the development of Internet technology,the traffic of video streaming in network traffic has been rising.Cisco claims in a report that streaming traffic transmitted through wireless Internet will account for 75% of total Internet flow in 2019 and it will keep increasing.Traditional streaming technology requires specialized Internet facilities,of which deployments and maintenance cost are high.Also,it disturbs the function of firewall.The wide application of Internet technology highlights the importance of dynamic adaptive video streaming based on HTTP protocol.Thus the MPEG-DASH technology was developed by MPEG in 2010,and then became the standard around the globe in November,2011.Gradually,it gained support from and was applied by major corporations.This thesis introduces the development and research of streaming technology,the system structure and file format of the DASH protocol.The study of rate adaptation algorithm for DASH has become a hot spot of research.The paper improves the classical rate adaptation algorithm.The decision to make requests towards segments with different qualities is made according to the buffer overflow level of the player and the network bandwidth,which makes it possible to reduce the switching times with segments of different bitrates in the cost of a little loss in average bitrate.At the same time,the average change of bitrate in switching becomes smaller.In addition,we transmit the DASH segments in our system using Server Push,which is the new feature of HTTP/2,decreasing transmission time and the number of HTTP requests towards the server.The buffer overflow level of the player can be perceived only in the playing terminal,and the segments being pushed to the client cannot be changed during the process so we set a secondary server based on the FDH/SAND mechanism to transmit the WebSocket message.Then the information about the buffer of client will be sent to the media serve terminal,so that the server can adjust the quality of DASH segments dynamically.It improves the bandwidth utilization ratio and makes the playing process more smoothly.Recently the advent of live industry has spotlighted the DASH live as a research focus.However,there are many differences between live streaming and video on demand.Low latency and preventing interruption of the video are two distinguishing features of live streaming.The latency of DASH live streaming is more than the length of one segment.Thus,we adopt chunk encoding to reduce live latency.Also,to improve the quality of video with same bitrate,we use HEVC instead of H.264 in original system.Live streaming requires the server to do real time encoding,packaging and segmentation after collecting enough data.While the adaptation feature of DASH needs segments with various bitrates and resolutions,it brings burden on encoding server.So we predict the request bitrate in advance,and encode the video according to predictive bitrate and the minimum bitrate with enough data.Decision algorithm decides to send segments with certain bitrate when requests are received by the server.The experiments prove that our scheme effectively reduces the load of server.