Design And Implementation Of Streaming Media System With Real-time Quality Evaluation Function

With the rapid development of the Internet,the emerging field of live streaming has also developed rapidly through Internet platforms.With the increasing business volume and people's increasing requirements for streaming media quality,network speed and network fluctuations have also brought great challenges to the live streaming industry.Therefore,streaming media quality assessment is not only important for user experience,but also for the development of streaming media live broadcast platforms.Streaming media quality assessment is divided into objective assessment and subjective assessment.The current streaming media quality assessment system only uses objective indicators such as bit rate,delay,and packet loss rate to perform objective assessment;subjective assessment requires a lot of manpower to score subjectively through subjective perception of streaming media quality in the same environment.Obviously,the quality assessment that requires the help of human resources cannot achieve an automated full-scale streaming media quality assessment function.In order to solve the above problems,this article has completed the development of a streaming media server based on the RTP protocol and an SDK for the docking module for the scene of enterprise conferences.On this basis,the function of real-time quality evaluation of streaming media is realized.In addition to the statistics of various objective quality indicators,this paper proposes an automated subjective quality evaluation strategy with the help of salient target detection.This strategy takes into account the human visual characteristics and the coding characteristics of H.264,uses the saliency map similarity between the blurred frame and the key frame to evaluate the impact of the blurred frame on the viewer's perception,and then outputs the subjective quality evaluation results.After testing,the method is similar to the traditional manual subjective quality assessment results.The saliency detection model used in this article has high real-time performance,and the SSIM loss value is added in the training,which makes the model more sensitive in brightness,contrast,structure,etc.After comparison experiments,it is found that the saliency detection model results using the optimized loss value more precise.This article also stores the streaming media quality evaluation data and completes the development of the administrator page,so that the administrator can view the real-time streaming media quality evaluation data.The realization of the real-time quality assessment function of streaming media enables the development team of the live streaming media system to analyze the current network conditions and service quality based on the quality assessment data,thereby further optimizing the system architecture and setting better media streaming parameters,etc.,to provide A better user experience.