Bayesian K-means clustering based quality of experience aware multimedia video streaming

Abstract

Media streaming is an essential approach for delivering multimedia information from the source distributor to the end-user through the Internet. Along with the development of more number of users and the spread of mobile devices, the availability and diversity of multimedia applications has also increased. Multimedia users primarily prioritize quality of experience (QoE), as they seek to access multimedia content with high availability and enjoy smooth video streaming in the shortest possible time. The impact of video delivery plays a significant role in QoE, which is efficiently made by delivering the content through a specialized content delivery network architecture. In this research, a Bayesian K-means clustering algorithm is proposed for the identification of QoE in multimedia video streaming. In this multimedia video streaming, the Amazon Prime video dataset is utilized for determining the performance of the proposed model. The proposed method is developed from the ‘Patching Up’ the video quality problem (PatchVQ) model, the from patches to pictures (PaQ-2-PiQ) model is utilized for the spatial feature extraction, and 3D ResNet-18 is utilized for temporal feature extraction. The proposed Bayesian K-means achieved a QoE reward function of 5,237.42 and 5841.36 as well as a fairness reward function of 5,841.36 and 8,732.08 at the speed of 1,500 kB/s and 2,000 kB/s respectively.