Research And Implementation Of Stream Media Server Based On Adaptive Bitrate

With the rapid development of communication technology and the Internet,network video services have become one of the major Internet applications.In order to ensure high-quality and reliable streaming media transmission,adaptive streaming media transmission technology came into being.The adaptive bitrate algorithm is the technical core of adaptive streaming media transmission,which effectively improves the users'Quality of Experience(QoE)by dynamically switching the video bitrate.Adaptive streaming media transmission technology has been widely used in various video-on-demand platforms.However,in the real-time live video streaming,the live broadcast platform is often accompanied by user communication and barrage interaction,and the latency has become one of the factors affecting the user experience.Therefore,in adaptive live streaming media transmission,the adaptive bitrate algorithm not only requires bitrate control,but also latency control.And there is relatively little information available for streaming media decision in live streaming media transmission.These issues have brought us new challenges.In adaptive live streaming media transmission,we can adjust the playback speed through continuous latency control parameters to reduce the live broadcast latency and reduce the re-buffering phenomenon.And we use the fragment skipping mechanism to prevent high latency.At present,the adaptive bitrate algorithms based on deep reinforcement learning(D-DASH,Pensieve)use DQN and A3C to design discrete bitrate control strategies,which does not consider continuous latency control.Although discrete control algorithms based on DQN can solve the discrete bitrate and continuous latency control problems by discretizing the latency control parameters,it will bring the problem that the discretization granularity is not well determined.In view of the above problems,this thesis proposes a continuous bitrate and latency control algorithm,and designs a live streaming media system based on this algorithm.In the bitrate and latency control,this thesis combines the DDPG algorithm,and designs continuous bitrate and latency control algorithm through feature state,network model,and action mapping,which effectively avoids the problem of discretization granularity,and provides more flexible and fine-grained decisions at the control level.In addition,the penalty weight of the re-buffering frequency is introduced into the reward function,so that the control algorithm converges in the direction of maximizing QoE,while also considering the effect of the re-buffering frequency.The effectiveness and applicability of the method are demonstrated through comparative experiments.In the design of the live streaming media system based on the bitrate and latency control,this thesis designs and implements the streaming media data processing system,the bitrate and latency control system,the client,and the streaming media monitoring system in detail.Finally,through the system function and performance tests,it is verified that the live streaming media system meets the expected goals.