Compressed-sensing-based video streaming in wireless multimedia sensor networks
MetadataShow full item record
Compressed sensing (CS) has emerged as a promising technique to jointly sense and compress sparse signals. One of the most promising applications of CS is compressive imaging. Leveraging the fact that images can be represented as approximately sparse signals in a transformed domain, CS allows images to be compressed and sampled simultaneously using low-complexity linear operations. Recently, these techniques have been extended beyond images to encode video. Much of the compression in traditional video encoding comes from using motion vectors to take advantage of the temporal correlation between adjacent frames. However, since calculating motion vectors is a significant source of power consumption, any technique appropriate for resource constrained video sensors must exploit temporal correlation through low complexity operations. In this dissertation, we develop a system that will allow high quality video streaming over wireless sensor networks (WSNs) to develop a wireless multimedia sensor networking (WMSN) device that is able to transmit video using battery operated, low power, low complexity, low cost devices. We will introduce the basics of video streaming over a WMSN, and examines what the challenges of this type of system are. First, we attempt to convince the reader why traditional WMSN systems are not practical for real applications, and why a new system must be developed. We then introduce a system architecture based on the properties of compressed sensing, that solves many of these problems by jointly designing a video encoder, a transport layer rate controller and a video streaming system. The goal of this system is to deliver high quality video while keeping within the constraints of a traditional scalar WSN. In the rest of this dissertation, we will examine three components of the CS based video streaming system. First, we will develop a compressive video sensing (CVS) video encoder that is able to encode video at very low complexity. We then develop a rate controller for both traditional video, along with video encoded using CVS, and show that, by taking video properties directly into account, we can develop a rate controller that optimizes the received video quality while maintaining fairness in terms of video quality. We then look at a system that corrects errors using a sparse error correction technique based on compressed sensing. We finally look at using relays to reduce the overhead required to correct errors.