Adaptive Bitrate Streaming Algorithm And QoE Research Based On Video Types

In recent years,with the increasing demand for streaming media,the way to improve video quality and the Quality of user's Qo E(Quality of Experience)in the limited bandwidth resources has become a research hot-spot.The purpose of the proposed adaptive bitrate streaming algorithm is to realize intelligent transmission control of streaming media,so as to make full use of the available bandwidth in the network and bring better streaming media video playback experience to users.However,the current algorithm ignores the effect of different video types on Qo E,which will make the algorithm to select low video quality or high delay for certain video types.These bad results will further affect the experience and also lead to waste of bandwidth.Therefore,this paper proposes the Qo E evaluation model for video types to optimize the adaptive bitrate streaming algorithm.The main innovations and works of this paper are as follows:Firstly,the development of the current DASH technology and the background of the three types of ABR algorithms are analyzed.According to the sensitivity difference of the Qo E indicators for the long,short and live videos,we extract the corresponding transmission parameters of the video type,and use the video classification factor to formulate the Qo E suitable for different videotypes.Further,we establish a model to analyze the changes of related parameters.The Actor-Critic algorithm is used to develop a bitrate adaptive algorithm which is suitable for the video type.By experiments,this method can improve the Qo E of Pensieve algorithm by 7% and the Qo E of MPC algorithm by 20%.Secondly,this paper analyzes the user's needs for transmission latency in live videos,analyzes the constituent factors of latency in live videos in detail,enumerates the work related to reducing live latency,incorporates latency into the Qo E evaluation system,DQN(Deep Q-learning)and A3C(Asynchronous advantage actor-critic)reinforcement learning methods are used to design and implement the corresponding low-latency ABR algorithm by controlling the cache.The performance of Pensieve in strong,weak and fluctuating networks is compared,given the loss of finite bit-rate score,the average delay is reduced by29%,69% and 68% respectively.In summary,this paper explores the transmission characteristics and needs of different video types,and formulates the Qo E of each type of video more accurately to drive the adaptive bitrate streaming algorithm.With the arrival and development of the era of artificial intelligence,training the network and formulating algorithms more pertinently are bound to be a general trend.The research results of this paper have certain academic significance and reference value for improving the quality of streaming video.