DefinePK

Abstract

Mobile Video Streaming is becoming increasingly popular in today’s Multimedia community. Various adaptive streaming techniques have been proposed by multimedia researchers to dynamically vary the video quality according to the available bandwidth. However, the deployment of best video adaptation techniques in real time is highly challenging due to critical QoE (Quality of Experience) requirements in wireless multimedia streaming. Resource constrained wireless multimedia networks demands better perception on the behavior of critical factors such as bandwidth in varying geographic milieu. In this paper, Machine Learning based online stream clustering is adopted to study the bandwidth impact in a streaming environment using 3G wireless video dataset. Massive Online Analysis (MOA) software framework is used to infer the results using algorithms such as CluStream and DenStream. The experimental result shows the effect of stream clustering based on unsupervised study. The measures such as Sum Square Error (SSQ) and Silhouette coefficient are deployed to perform cluster analysis. The results demonstrate the efficiency of CluStream with K means algorithm over density based streaming algorithm. The proposed framework justifies the scope of context aware computing applications in the broader areas of wireless multimedia.