CS: Robustness of Compressed Sensing in Sensor Networks

Here is a fresh undergraduate thesis by Brett Hern, a student at Texas A&M University, entitled Robustness of Compressed Sensing in Sensor Networks. The abstract reads:

Compressed sensing is a new theory that is based on the fact that many natural images can be sparsely represented in an orthonormal wavelet basis. This theory holds valuable implications for wireless sensor networks because power and bandwidth are limited resources. Applying the theory of compressed sensing to the sensor network data recovery problem, we describe a measurement scheme by which sensor network data can be compressively sampled and reconstructed. Then we analyze the robustness of this scheme to channel noise and fading coefficient estimation error. We demonstrate empirically that compressed sensing can produce significant gains for sensor network data recovery in both ideal and noisy environments.

It can be downloaded here.


Image credit: NASA/JPL-Caltech/University of Arizona/Texas A&M, Martian Polar Plains taken by the Phoenix's Surface Stereo Imager.