Research On Adaptive Transmission And Model Adaptation In Mobile Video System

With the development of communication and Internet technology,there is an increasing demand for video applications.In order to actively respond to this trend,media providers have adopted various methods to improve the quality of user experience,among which the adaptive bitrate(ABR)algorithm has attracted much attention.As the mainstream business scenario,the mobile terminal needs a tailor-made ABR algorithm because the network is more unstable and there are special mobile events such as base station switching.Although there are many studies on the ABR algorithm,the method designed for this scenario is relatively fewer.To this end,this paper firstly proposes a payload-aware ABR scheme PAR,which alleviates the insufficient bandwidth prediction ability of the traditional ABR algorithm in fluctuating networks and accurately assist decision-making.In order to improve the generalization of data-driven algorithms,PAR also introduces an online learning framework based on bandwidth distribution change point detection,and adjusts the model in a timely manner to cope with rapid changes in the network.Furthermore,in order to adapt to the mobile user scenarios,this paper also proposes a handover-aware based code rate decision post-processing method HB-FST,through the finite state transition machine model,to actively avoid the physical layer base station handover event from the application layer ABR decision.A large number of experiments based on real network data shows that in terms of user QoE indicators,PAR can exceed classic baseline methods by 2.66%to 79.43%on different datasets.In the three mobile scenarios of high-speed rail,driving and walking,HB-FST has significantly optimized multiple performance indicators of the original ABR algorithm,and the highest quality playback ratio has increased by nearly 2 times.