The Study And Implementation Of Dynamic Streaming MPEG-DASH

With the development of streaming technologies, the demand for streaming is gradually increasing, and for this reason, the streaming technologies based on HTTP is leading the trend. There are various disadvantages for traditional streaming technologies. For example, it cannot deploy the existing network structure and it has some problem when it comes to the traversal of the firewall. HTTP progressive download will create a bad influence on use experience under the frequently changed network. For the reasons above, dynamic adaptive streaming has been a key technology for streaming. For now there are three mature adaptive streaming technologies in industry area, which are Apple HTTP Live Streaming, Microsoft Smooth Streaming and Adobe HTTP Dynamic Streaming. MPEG and 3GPP begin to draw the protocols on April 2014 and name it DASH. This protocol become international reference on April 2012.In this paper, we introduce the development of Dynamic Adaptive Streaming, the ratification and implementation of DASH protocol. We also elaborate the data model of MPD and the format of segment. The most important part of MPEG DASH is the rate adaptation algorithm. In this paper, we study the classic algorithms for rate adaptation, and sum up the advantages and disadvantages of these algorithms. At last, we propose a novel algorithm based on fixed-interval buffer model. We set a fixed length buffer interval. The proposed algorithm can compensate bandwidth fluctuation, thus reducing switching times. Combined with the proposed algorithm, we also propose an intermediate quality level algorithm, which can improve subjective QoE. In addition, we propose a quick boot algorithm, which can select the proper representation as soon as possible. With the comparison between the proposed algorithm and state-of-art algorithm under the same experimental environment, the proposed algorithm can dramatically decrease the number of bitrate switches with the similar average bitrates, thus enhancing the user experience.Currently, the encoders for DASH and the platforms for DASH are not mature enough. Under that circumstances, we design a new DASH Encoder in this paper, which can support H.264 and HEVC and support both client for ITEC and GPAC.At last, we design a cloud encoder for DASH, which can support the functions that users can upload initial videos and the cloud encoder can encode the video into DASH materials from the server side, and then provide the users the final DASH files to download or provide the users a URL which can play the DASH video. The cloud encoder can solve the problem that the DASH dataset is under shortage and can enable the researchers to make DASH materials by themselves conveniently.In summary, we have made several innovations in both server side and client side, which can perform well under the proof of rigorous experiments.