Research On Panoramic Video Transmission Based On Viewport Prediction

Due to the continuous innovation of computer science and multimedia technology,VR(Virtual Reality)has received more and more attention.As an important part of VR technology,panoramic video has also been applied in the fields of leisure,education and medicine.However,the large amount of panoramic video and delay-sensitive features have put tremendous pressure on existing transmission networks.Therefore,how to maximize the viewing experience of multiple users under a limited network bandwidth is crucial.This paper mainly optimizes the transmission of panoramic video from the perspective of viewport prediction and physical resource scheduling.In recent years,more and more researches have been made on the viewport-based adaptive transmission scheme for panoramic video.The key premise of this scheme is the accurate prediction of the viewport position.This paper improves the viewport prediction algorithm based on the historical viewing trajectory of the user.Firstly,based on the historical viewport data of the current user,we use the LSTM(Long-Short Term Memory)model to initially predict the coordinates of the center of the viewport.Then combine the viewing trajectory of other users to adjust the preliminary forecast results.The experimental results show that the accuracy of the proposed prediction algorithm is improved compared with the benchmark algorithm.On the other hand,how to scientifically schedule bandwidth resources to meet the needs of users of different services such as panoramic video and ordinary video is also a research focus.This paper first studies the classic downlink resource scheduling algorithms in the LTE(Long Term Evolution)system,and then proposes a scheduling algorithm based on Q-Learning for the mixed service scenario that introduces panoramic video services to meet the user’s QoS(Quality of Service).The algorithm can adaptively select the most suitable scheduling strategy in each scheduling interval according to the network status.Simulation results show that compared with the classic scheduling algorithms,the improved algorithm can effectively reduce the packet loss rate and delay of the panoramic video service,and improve the throughput of real-time services in a mixed service scenario.