WVSNP-DASH: Name-Based Segmented Video Streaming

摘要

Video streaming from sensors and miniaturized devices is attractive for a wide range of web-based applications, e.g., remote surveillance. Existing web-based video streaming frameworks, such as the hypertext transfer protocol (HTTP) live streaming (HLS) and the motion picture experts group’s dynamic adaptive streaming over HTTP (MPEG-DASH), have dependencies between the individual video segments and a manifest file that contains video metadata. Also, existing web-based video players are limited to fetching video segments over TCP/IP networks. The video segment dependencies complicate video segment distribution by resource-constrained source nodes, which may employ non-TCP/IP protocols, such as Zigbee. This paper proposes and evaluates a wireless video sensor network platform compatible DASH (WVSNP-DASH) framework and a WVSNP-DASH player (WDP) for flexible web-based access of video from sensors and other miniaturized source nodes. The WVSNP-DASH framework is based on independently playable video segments with a specific naming syntax that conveys elementary metadata so as to facilitate flexible search, transfer, distribution, and playback. The WDP employs elementary processes of version 5 of the hypertext markup language (HTML5) for video buffering and playback. Video segments are fetched into the HTML5 file system space, permitting flexible video fetching over a wide range of protocols, including sensor network protocols. Comparative evaluations of a WDP prototype with optimized HLS and MPEG-DASH players indicate that WDP has low client (receiver) load, while providing significant potential for power savings on the source node serving the video streams.