Network-compressive coding for wireless sensors with correlated data

Ketan Rajawat, Alfonso Cano, Georgios B. Giannakis

Research output: Contribution to journalArticlepeer-review

8 Scopus citations

Abstract

A network-compressive transmission protocol is developed in which correlated sensor observations belonging to a finite alphabet are linearly combined as they traverse the network on their way to a sink node. Statistical dependencies are modeled using factor graphs. The sum-product algorithm is run under different modeling assumptions to estimate the maximum a posteriori set of observations given the compressed measurements at the sink node. Error exponents are derived for cyclic and acyclic factor graphs using the method of types, showing that observations can be recovered with arbitrarily low probability of error as the network size grows. Simulated tests corroborate the theoretical claims.

Original languageEnglish (US)
Article number6353398
Pages (from-to)4264-4274
Number of pages11
JournalIEEE Transactions on Wireless Communications
Volume11
Issue number12
DOIs
StatePublished - Nov 29 2012

Keywords

  • Network coding
  • compression
  • graphical models
  • wireless sensor networks

Fingerprint Dive into the research topics of 'Network-compressive coding for wireless sensors with correlated data'. Together they form a unique fingerprint.

Cite this