A Research On Interactive Streaming Media Transmission Strategy And Experience Optimization

As on-demand evolves to live streaming,virtual reality to remote surgery,and autonomous driving to industrial internet,the application of streaming media has evolved from human-machine face-to-face interaction to fullscenario immersive interaction,and then to industrial-landed interaction between humans,machines,and physical objects.The number of devices connected to the network is increasing,and the frequency of streaming media interaction and transmission volume is growing rapidly.As the demands of interactive streaming media are mapped from application scenarios to new transmission requirements,the requirements for network latency and user experience will become higher.At the same time,with the basic deployment of 5G base stations completed,the increase in network nodes will begin to slow down,and the reduction in core network transmission latency will become increasingly expensive.Network service providers can no longer reduce endto-end transmission latency by increasing network nodes.How to use edge servers and design excellent last-mile transmission strategies to meet personalized user experience requirements has become an important problem that needs to be solved urgently.Currently,HTTP-based adaptive streaming media transmission protocols have dominated the market,and the core idea is to divide streaming media content into multiple media chunks with a cycle of a few seconds and transmit them one by one.It can dynamically adjust the media bitrate according to the network condition and device capability,and users can freely drag and switch media streams.However,its transmission cycle is at least two media chunks long and cannot guarantee ultra-low-latency media transmission and flexible bitrate switching.The adaptive bitrate selection algorithm considers high bitrate,low rebuffering,and high smoothness indicators,and can adaptively select media bitrate based on current network status,cache size,video content,and user preferences to provide the best user experience.However,specific to emerging interactive streaming media applications,considering low latency and low-cost requirements,further exploration is needed.This article summarizes the research work on designing and using interactive streaming media transmission strategies to achieve four performance aspects:"ultra-low latency","accurate switching","secure and reliable",and "deterministic latency".The corresponding achievements obtained are as follows:(1)A frame-based interactive streaming media transmission strategy is proposed.On the server side,a reinforcement learning-based adaptive bitrate selection algorithm is designed to adaptively select the bitrate of future fr-ames based on observable state information.It cleverly solves the problem of long waiting time between bitrate decision and bitrate execution caused by misaligned I-frames and the problem of mismatch between action and reward.As the reinforcement learning algorithm tends to be conservative during the training process in weak networks,a rule-based adaptive bitrate selection algorithm is designed specifically for weak networks as an optimization.On the client side,three latency control mechanisms are designed to achieve framebased fine-grained control:slow playback,fast playback,and frame skipping.Compared with traditional methods,this research meets the user demand for ultra-low latency in interactive streaming media.The average latency is reduced by 32%-77%,and the average user experience is improved by 28%67%.(2)A variable-length video chunk-based bitrate switching interactive streaming media transmission strategy is proposed.After carefully studying the impact of video chunk cycle length on user experience,a streaming media transmission protocol based on an intelligent triggering mechanism is designed to balance the timeliness of bitrate switching and system resource consumption.A bitrate adaptive selection algorithm based on data-driven I-frame prediction is designed to predict the switchable bitrate within the next physical decision cycle,maximizing the accuracy of bitrate switching.Compared to traditional methods,this research meets the user demand for accurate switching in interactive streaming media,reduces the average bandwidth overhead by 37%57%,and improves the average user experience by 15%-49%.(3)An interactive streaming media transmission strategy is proposed for precise detection of new anomaly classes.As network traffic comes from the complex interaction between users and social applications,a recurrent neural network-based autoencoder is designed to extract temporal features,and a cumulative importance sorting algorithm is designed to extract statistical features.A meta-learning-based multi-class anomaly detection algorithm is designed to accurately detect small samples,imbalanced,self-similar,and small-labeled new anomaly classes.Compared to traditional methods that passively detect specific abnormal classes using a large number of labeled training samples,this research meets the user demand for secure and reliable in interactive streaming media,and the average F1-score is increased by 12%45%.(4)A delivery rate as the goal interactive streaming transmission strategy is proposed.The video stream is divided into multiple media elements,and users are given personalized latency sensitivity thresholds and priorities.Based on the Inflight bandwidth prediction model and the scheduling algorithm based on personalized latency sensitivity thresholds and media element priorities,more video chunks in multiple links can be delivered completely within deterministic latency(delivery rate).Compared to traditional methods,this research abandons the goal of maximizing throughput and ultra-low latency,and meets the user demand for deterministic latency in interactive streaming media,and the average user experience is improved by 12%-31%.