Researches On Adaptive Transmission For 360-degree Video Streaming

In recent years,360-degree videos,which bring immersive experience for users,have become increasingly popular.However,delivering the 360-degree video requires ultra-high bandwidth over the internet and the network infrastructure nowadays hardly meets the requirement.To save bandwidth for the 360-degree video streaming over the internet,in this paper,we propose a transmission strategy for tile-based adaptive 360-degree video streaming,which is able to maximize the quality of experience(QoE)of viewers under the limited and time-varying network conditions.Specifically,we first derive the accurate predicted viewing probability of any given spherical point from the known head rotation prediction of users.Having known the predicted head rotation of the viewer in advance,we fit the probability density function of prediction error which is prediction window-dependent by numerical simulations.Following the function,we calculate the viewing probability of any given spherical point and any given tile also.Such viewing probabilities are applied in calculating the prediction error-dependent viewport which can avoid the mismatches between the predicted viewports and the real viewports of the viewer.At the basis of viewing probability,we then propose an adaptive tiling scheme able to achieve a good performance on the bandwidth utilization during the video transmission.The proposed scheme aims to adaptively select the proper granularity of tiles at different position of a given frame in order to reduce the redundant information resulting not only from the encoding processing but also from the bit allocation processing.The results have shown that our proposed scheme requires the lowest bitrate for the video streaming file under the same bitrate allocations.In respect of the adaptive bitrate algorithm for the 360-degree video streaming,a QoE metric and optimization problem are firstly formulated in order to maximize the long-term QoE of the user by adapting the quality of tiles to the time-varying network conditions.We further model the rate adaptation logic for the optimization problem as a Markov decision process and employ the DRL-based algorithm to dynamically learn the optimal policy that guides the selection of the tiles' bitrate.Simulation results show the superior performance of the proposed method compared to the existing methods.