The Research And Implementation Of Tile-based Driven By User Data

With the continuous development of network communication technology and video services,the new generation of multimedia services represented by virtual reality has become one of the most popular services on the Internet.Panoramic video,as an implementation technology of virtual reality,has also received wide attention from academia and industry in recent years.However,the high bandwidth,high storage requirements,and latency-sensitive characteristics of panoramic video technology have posed great challenges to current multimedia service providers.The thesis conducts technical research on Tile-based panoramic video latency optimization for the most common panoramic video technology,which is Tile-based panoramic video with Equiangular Cubemap projection method.The main contents are as follows.Based on the data-driven cloud-edge collaborative caching strategy,the thesis uses an improved spherical DBSCAN clustering algorithm to extract user region of interest(ROI)features.The ROI features are innovatively introduced into the edge caching scheme for panoramic video,and the binary particle algorithm is used to solve the edge storage decision based on Tile video blocks under the cache capacity constraint,thus improving the cache utilization.The thesis starts with edge caching technology to reduce service latency by reducing the delay and communication overhead caused by intermediate transmission in the core network.The adaptive playback strategy based on viewport prediction first uses Gaussian process regression with ROI features to predict the future viewport range of users.To solve the problem of inaccurate field-of-view prediction due to user’s random behavior,the thesis improves the existing player by proposing a new re-downloading frame chasing strategy with field-of-view compensation to replace the low-definition Tile video blocks in the user’s field of view in a timely manner.The thesis starts with preloading technology to pre-fetch the video resources that users may watch.thus further reducing the transmission delay.Tile-based panoramic video transmission system in order to further validate the proposed method,the thesis design and implements an end-to-end panoramic video transmission system.Among them,the content provider side is responsible for processing and storing the panoramic video,and also includes the user data analysis function;the edge server side undertakes the work of collaborative caching and content scheduling;the client side is responsible for the tasks of sending panoramic video requests and experimental effect verification.Based on this system,the performance of the proposed method is verified.In summary,the thesis takes the Tile-based panoramic video and EAC projection method as the research object and proposes methods to reduce the latency from the panoramic video transmission stage and panoramic video playback stage by analyzing the time consumption of the whole link process of panoramic video viewing,aiming to optimize the overall latency of Tile-based panoramic video and thus effectively improve the user experience.