Improving 360-Degree Video Field-of-View Prediction and Edge Caching

Aliehsan Samiei, Ravi Prakash

Abstract: Streaming 360° videos over the Internet to Head Mounted Displays (HMDs) requires high bandwidth owing to their high resolution. Viewers, limited by the Field of View (FoV) of the HMD, can only watch a small portion of the scene at a time. Therefore, predicting the future FoV of viewers and caching popular content in the mobile edge cloud can reduce both downloading latency and traffic over the core network. The primary contributions of this work are: (i) a new strategy for generation of viewing heatmap which significantly improves our understanding of regions of viewer interest, and (ii) demonstrating the relevance of some indices used in economics and election analysis to further narrow down the viewing hot spots that can be most useful in making caching decisions and FoV prediction. We demonstrate how these indices and the improved heatmap can improve caching and FoV prediction for a number of videos drawn from two independent datasets. Most of the computation needed to implement these concepts can be performed offline by video content servers using previous viewers’ FoV traces. Hence, the computation load on viewer devices is also low.