Compressed sensing of Gauss-Markov random field with wireless sensor networks

We propose a scalable and energy efficient method for reconstructing a dasiasparsepsila Gauss-Markov random field that is observed by an array of sensors and described over wireless channels to a fusion center. The encoder is universal, i.e. invariant to the statistical model of the source and the channel, and is based on compressed sensing. The reconstruction algorithms exploit the a-priori statistical information about the field and the channel at the fusion center to yield a performance comparable to information theoretic bounds. Furthermore, by putting stringent constraints on the sensing matrix we avoid (or even eliminate) inter-sensor communication while suffering negligible degradation in performance.