Sparse ConvNet and Policy Driven Encoder for Enhancing Cloud-based QoE in 5G and Beyond Networks

Abstract

Optimizing video streaming quality for extended-duration videos in cloud-based Quality of Experience (QoE) presents significant challenges due to the dynamic nature of network conditions, user engagement patterns, and resource allocation demands. Hence, a novel Sparse ConvNet and Policy Driven Encoder for Enhancing Cloud-based QoE in 5G and Beyond Networks is proposed for optimizing video streaming quality and resource allocation in real-time. Existing neural networks struggle fail to solve this data traffic issue, which increases computational burden and degraded QoE for users, this leads to customer’s dissatisfaction. Thus, Sparse Graph Attention ConvNet (SGA ConvNet) is introduced to analyze the data traffic. This reduces computational complexity while optimizing network resources in real-time, ensuring a balance between operational costs and QoE. Moreover, traditional video encoding algorithms struggle with rapid scene changes and high motion, which increase data volume and computational intensity, reducing overall efficiency. To address this, Policy-Driven Variational Encoder (PD-VE) is introduced, enhancing adaptability to network conditions and improving the viewer's streaming experience by alleviating congestion and latency issues. The suggested method outperforms existing techniques in optimizing video streaming quality and resource allocation, achieving higher accuracy, precision, while minimizing RMSE and encoding time, as demonstrated by experimental results.