Design And Implementation Of Congestion Control System For Video Communication

With the development of the Internet,video transmission accounts for an increasing proportion in network transmission,and video has become a more and more important part of network life,including video conference,short video platform,live broadcast and other scenes.Due to the uncertainty of network stability,it is necessary to use congestion control algorithm to reduce the impact of network uncertainty on video transmission.The traditional congestion control algorithm is difficult to meet the requirements of low delay and tolerance of a certain packet loss rate in real-time video transmission in video call scenario.In recent years,some congestion control algorithms for video transmission can be improved in terms of comprehensive performance.With the rise of artificial intelligence algorithms,reinforcement learning and other technologies have also been tried to be applied to the implementation of congestion control algorithms,but these are only academic attempts,and the current practical application is still dominated by traditional methods.By building a simulation platform,this paper compares the SCReAM algorithm with the GCC algorithm,and verifies that the SCReAM algorithm has the problem of low network bandwidth utilization in the simulation,but the delay has certain advantages compared with the GCC algorithm.This paper proposes K-SCReAM algorithm to try to solve the congestion control problem in real-time video transmission:first,improve bandwidth utilization,then estimate the delay in the network at the sender by using Kalman filter,and use the predicted delay to adjust the sliding window to reduce the occurrence of congestion.Secondly,use error control technology to optimize the data recovery and transmission rate adjustment after packet loss according to the scene of real-time video transmission,so as to better adapt to the network situation and improve the user experience when packet loss.In addition,developing the video communication system which supports cross platform,and the client is modularized,which provides a simpler method for integration in other software and more platforms.