Spectrum efficient coding scheme for correlated non-binary sources in wireless sensor networks

Energy-aware technique to reduce energy consumption in distributed sensor networks has become a prominent topic in sensor network research. Various sensor network applications have taken energy efficiency into consideration, in the case of correlated binary sources, distributed source coding has been literally studied in information theory. However, data sources from real sensor networks are normally non-binary. In this paper, we proposed a spectrum efficient coding scheme for correlated non-binary sources in sensor networks. Our approach constructs the codeword cosets for the interested source, taking advantage of statistical characters of the distinct observations from sensor nodes. The coset leaders are then transmitted via the channel and decoding is performed with the available side information. Simulations are carried out over independent and identically distributed (i.i.d) Gaussian sources and data collected from Xbow wireless sensor network test bed. Simulation results show that the proposed scheme performs at 0.5 - 1.5 dB from the Wyner-Ziv distortion bound.