While there has been progress in recent years in solving numerous wireless sensor networking challenges, the key problem of enabling real-time quality-aware video streaming in large-scale wireless networks of resource-constrained devices is still open and largely unexplored. Existing wireless networking protocol stacks based on transmitting predictively-encoded video are computationally expensive, have limited resilience to wireless channel errors, and use available network resources inefficiently. This project is attempting a new approach based: (a) On the development of a novel wireless streaming framework for resource-constrained devices rooted in the theory of compressed sensing (CS) and (b) co-design/optimization of the video encoder and key wireless networking functionalities.
The new networked wireless streaming system being developed is referred to as Compressive Video Streaming (CVS) and has the potential to significantly reduce power consumption for a given target video quality for resource-constrained sensing devices. New network control algorithms are designed and integrated with the video encoder that, unlike TCP and TCP-friendly approaches, use the estimated received video quality as the basis for resource allocation decisions. Project implementation and extensive testing is being carried out on several experimental platforms.
The technology to be developed has the potential to strongly impact the state of the art in resource-constrained wireless video sensor networks. Important educational objectives of the program include mentoring of minority graduate and undergraduate students, and development of new interdisciplinary course materials.