Research On Video Transmission QoS Optimization Scheme

The rapid development of computer and communication technology has promoted the development of network video technology.Common applications include remote video conferencing,video surveillance,live video,and so on.However,the existing IP network transmission does not provide video quality assurance.Video data is lost when the network environment is very poor.For scenes with high real-time video conferencing and live broadcast,the user experience is extremely poor,with reduced video QoS.Therefore,research on how to improve video QoS is important and urgently needed.This thesis mainly improves the QoS of video by studying the video transmission in congestion control and video playback control.Firstly,the implementation principles and application scenarios of BP network,RNN network,LSTM network and Markov model are studied,laying a theoretical foundation for the proposed algorithm.Secondly,considering the time series characteristic of Internet traffic data.A congestion control algorithm predicting Internet bandwidth utilization based on LSTM is proposed to alleviate network congestion during video transmission.The core idea that the transmission rate of the video is dynamically adjusted according to the bandwidth utilization of the Internet predicted by using LSTM.This thesis compares BP network with LSTM network on the accuracy of Internet bandwidth utilization prediction.The experimental results show that the LSTM network has higher accuracy than the BP network.The algorithm can accurately grasp the Internet changes,achieving good results for congestion control,and improve the QoS of video transmission to some extent.Finally,this thesis proposes a playback control algorithm based on Markov chain,aiming at solving the shortcomings of existing playback control algorithms.The core idea of the algorithm is dividing the receiving buffer into three states: upper overflow state,flat steady state and underflow state.Inputting the state at the previous moment into Markov model,the state transition matrix is calculated to predict the state at the next moment of the buffer.The playback rate of the video is dynamically adjusted according to the predicted state.The proposed algorithm is verified on the experimental environment built on the simulation platform NS2.The results show that the playback control algorithm proposed in this thesis has less fluctuation in video playback rate than other algorithms.Adjusting the playback rate of video according to the actual state of the buffer reduces the average playback delay of the video,and improves the QoS of the video to some extent.