Design And Implementation Of Anti-Packet Loss Audio Processing System In Video Conference

With the development of Internet,multimedia real-time communication technology has brought us a lot of convenience in life.As a mainstream solution to audio and video application,such as video conferencing,Taobao live broadcast and live broadcast of events on various platforms,WebRTC transmits multimedia streams based on the UDP protocol.UDP is a best-effort connectionless transmission technology which does not have voice quality guarantee.Therefore,packets may be degraded by network impairments,such as packet loss and packet burst.The audio processing technology plays an important role in resisting these network impairments.In this thesis,an anti-packet loss audio processing system is designed and developed based on WebRTC which is used in video conferencing.In order to understand the scenarios of packet loss,a video conferencing platform based on WebRTC is built in this thesis.This platform is used to record and observe the probability distribution of packet loss rate under WIFI network.And then,identify causes of packet loss in audio packet burst scenario.In this work,we improve the anti-packet loss capability of the audio processing system from two aspects:First,we replace the packet loss concealment scheme of the original WebRTC with a hybrid model based on LPCNet,which improves the packet loss concealment capability.Second,we design and implement an improved jitter buffer algorithm based on original WebRTC.The algorithm is used to the speech quality distortions caused by audio packets in bursts.Silence detection is used to reduce the impact of packet loss in the algorithm.Finally,we validate that the system improves the capability to resist packet loss compared to original WebRTC audio processing solution,and meets the design requirements of the system.