Research And Implementation Of Dynamic Adaptive Bitrate Simulation Platform Base On Live Video Scene

In recent years,with the rapid development of Internet technology and network video business,video users have increased rapidly.Among them,the number of the live video users has been an explosive growth because.of its real-time,interactive and other characteristics.Due to the complexity of the network environment and the diversity of live streaming scenes,an adaptive bitrate algorithm which can effectively improve the quality of users’ video viewing experience has become a research hotspot in streaming media transmission.However,the live streaming architecture is complex and cumbersome,so it is difficult to train and evaluate the algorithm quickly in the real scene A simulator which can quickly and accurately simulate the real-time adaptive streaming environment is very important for the research of bitrate algorithm.For this reason,this paper designs and implements a dynamic adaptive bitrate simulation platform based on live video scene.It provides one-stop service for algorithm researchers in related fields from data generation,environmental simulation to algorithm evaluation.In this paper,the simulation platform is divided into data generation module,simulation module and auxiliary module.Through the data generation module,users can generate sufficient network and video data according to any business scenario;after inputting the data into the simulation module,the key indicators that may affect bitrate decision-making can be obtained from the output of the live streaming simulator with delay control function to design and train the algorithm.Finally,according to the user QoE model provided by the auxiliary module,a comparative experiment with the traditional algorithm is carried out and evaluated according to the visual information graph of the results.The business process test and function test of the system show that the realization of the simulation platform meets the expected goal.In addition,based on the simulation platform,this paper uses the reinforcement learning proximal policy optimization algorithm combined with the generative adversarial network to optimize the adaptive bitrate algorithm,and to verify the availability of the platform and the effectiveness of the generation adversarial network to solve the key information loss problem of reinforcement learning.