Research On Jitter Buffer In Multimedia Data Transmission Based On Regression Model

The integration of Internet and communication technology and the widespread popularity of mobile intelligent terminals provides a powerful network transmission and hardware support for multimedia real-time communication(RTC)technology.Multimedia RTC services are gradually spreading throughout people's work and life,such as video chat,webcast,online education and so on.Jitter buffer plays a key role in compensating for jitter,smoothing data transmission,and improving playback quality.It is one of the important modules affecting service quality.There are type of two jitter buffer now: static jitter buffer and dynamic jitter buffer.The static jitter buffer uses a fixed depth buffer setting during communication and cannot cope with different network environments.The existing dynamic jitter buffer either sets the buffer thresholds,triggers adjustment when the threshold is reached,or adjusts by predicting network traffic.These algorithms all adjust lately and cannot describe network changes or compensate for jitter accurately.This paper deeply analyzes the principle and workflow of the dynamic jitter buffer algorithm in WebRTC(Web Real-Time Communication).On this basis,using a new filtering method to estimate the network queuing delay and calculate the jitter delay reference value around the key problem of jitter delay effect.The method introduces a machine learning regression model to predict the packet loss rate and end-to-end delay of the system based on network conditions and jitter delay.The optimal jitter delay is searched for near the reference value according to the prediction and updated into the video frame.In order to evaluate the effect of the jitter buffer module,this paper establishes a transmission quality assessment model based on QoS(Quality of Service)parameters.The model takes into account the system's packet loss rate,end-to-end delay and playback jitter.Under the same network environment and media processing engine,the transmission quality score can be used as feedback for the working effect of the jitter buffer module.Finally,a comparison experiment with existing jitter buffering algorithms in WebRTC under different network environments is designed.After testing,the jitter buffering algorithm based on machine learning prediction model can be obtained with less delay cost in different network environments.Better network jitter compensates for the effect.