Research On Video Streaming QoE Based On The Combination Of Network Parameters And Video Quality

With the development of video technology and network communication technology,video services are showing an explosive growth trend,and users' requirements for video services are also increasing.Accurately grasping the Quality of Experience(QoE)when users watch videos can effectively provide guidance for resource allocation,service improvement,and new business research and development of video services.In order to better serve users,video service providers and operators are paying more and more attention to how to evaluate the quality of video user experience.Regarding the evaluation of video QoE,most of the current research angles are single and cannot fully reflect the characteristics of video QoE,and the nonlinearity and complex time dependence involved in video QoE are not considered.In view of the above-mentioned research status,this article comprehensively researches video QoE from two perspectives: network Quality of Service(Qo S)parameters and Video Quality Assessment(VQA).The main contents include:(1)Based on three public data sets,first analyze 11 Qo S parameters and 3 VQA methods,then combine the Qo S parameters representing the network impact and the VQA representing the human perception impact,and finally perform modeling on the data set verification.The experimental results show that when the Qo S index buffer and the structural index score of the VQA method are used as input,the prediction result of the model has the highest consistency with the subjective evaluation.The SRCC of the model is 0.87,which is 8.75% higher than the structural index,indicating the buffer and structure.The index can be used as a key influencing factor in the construction of a video QoE evaluation model,providing a reference basis for real-time adjustment of playback fluency of video services and requesting to download video clips,thereby improving the quality of user experience.(2)Propose an improved video QoE prediction scheme.This scheme first introduces an attention mechanism in the long-short-term memory neural network;secondly,it uses the Mix Up method to amplify the number of samples in the data set and increase the sample diversity;finally,the buffer and structural index are used as input,and it is performed on two public data sets.Experimental verification.The experimental results show that the improved video QoE prediction scheme on the SQoE-I data set and SQoE-III data set,PLCC is 0.95,0.88,SRCC is 0.94,0.87,respectively,better than other advanced models,and has better subjective evaluation consistency.Explains that the improved solution can handle the nonlinear interaction between network parameters and video quality in video QoE,capture the complex time dependence involved in video QoE,and accurately predict video QoE,which is more suitable for frequent video quality switching than research in a single direction,Adaptive video transmission scenarios with fast bandwidth changes can be used as evaluation criteria for video services.(3)In the 5G scenario,there is a lack of comprehensive analysis of bit rate adaptive algorithms and video QoE.This article conducts video adaptive transmission experiments in 5G mobile and static scenarios,analyzes the relationship between network parameters,video QoE and bit rate adaptive algorithms,and uses video QoE to select appropriate bit rate adaptive algorithms to improve user experience quality.The results show that the user experience quality can be improved up to 48.67% in the mobile scene,and the user experience quality can be improved up to 121.12% in the static scene.