Key Technologies Of QoE Optimization For Video Conference

Video conference is increasingly being adopted by end consumers.Due to the need of Real-time communication(RTC),it is extremely challenging to deliver good video conference QoE(quality of experience) over wireless networks.To achieve good realtime QoE performance, RTC applications require to adapt to network performance.In this paper,we focused on adaptation scheme for RTC applications and networkperformance prediction strategies.The Main Content is as follows:1.In order to get a good understand of how video conference sofwares improve their QoE,we conduct a measurement study on three popular mobile video call pplications: Face-Time and Skype, over both Wi Fi and Cellular links.Packet traces are captured at different points using Wireshark and tcpdump.With trace analysis,we obtain valuable insights regarding disadvantage of exsting solutions that they do not make full use of network performance.We also analyzed possible QoE degradation in Wireless environment.2.Assuming we know the bandwidth for the entire video.we obtained the optimal uitilization model of bandwidth.We observed that existing algorithms failed to take full advantage of avalible in video transmission.Relaxing the assumption to knowing the avaible bandwidth a few seconds into the future, we designed SPA algorithm.we observe that prediction alone is not sufficient.However,when combined with rate stablilization functions, SPA evolve into PBA algorithm,which ourperforms existing algortim as signigicantly as 40%. Our results lead us to be lieve that we can tremendously improve video QoE by predicting available bandwidth.3.Proved that forecasting short-term packetloss, delay and available bandwidth in cellular nerworks is possible due to its channel estimation scheme and resource scheduling algorithm.We develop a algorithm called PROPHET,which passively collects current network performance, and then uses regression trees to forecast future network performance. PROPHET successfully predicts the occurrence of packet loss with in 0.5s time window for 97% accuracy and occurence of long one-way delay for 98% accuracy and thoughroughput with less than 10 kbps error.In emulation, we increase the QoE of a video conferencing by up to 13 dB using PSNR scale.