Design of optimal convolutional codes for joint decoding of correlated sources in wireless sensor networks

We consider a wireless sensors network scenario where two nodes detect correlated sources and deliver them to a central collector via a wireless link. Differently from the Slepian-Wolf approach to distributed source coding, in the proposed scenario the sensing nodes do not perform any precompression of the sensed data. Original data are instead independently encoded by means of low-complexity convolutional codes. The decoder performs joint decoding with the aim of exploiting the inherent correlation between the transmitted sources. Complexity at the decoder is kept low thanks to the use of an iterative joint decoding scheme, where the output of each decoder is fed to the other decoder’s input as a-priori information. For such scheme, we derive a novel analytical framework for evaluating an upper bound of joint-detection packet error probability and for deriving the optimum coding scheme. Experimental results confirm the validity of the analytical framework, and show that recursive codes allow a noticeable performance gain with respect to non-recursive coding schemes. Moreover, the proposed recursive coding scheme allows to approach the ideal Slepian-Wolf scheme performance in AWGN channel, and to clearly outperform it over fading channels on account of diversity gain due to correlation of information.