Joint source-channel decoding for LDPC-coded error-corrupted binary Markov sources

We consider the problem of joint decoding and data fusion in data gathering for densely deployed sensor networks modeled by the Chief Executive Officer (CEO) problem. More specifically, we consider the binary CEO problem where all sensors observe the same time-correlated binary Markov source corrupted by independent binary noises. Hence, the observations are two-dimensionally (temporary and spatially) correlated. In the proposed scheme, every sensor apply a low-density parity-check (LDPC) code and transmit the corresponding codeword independently over additive white Gaussian noise (AWGN) channels. To reconstruct the original bit sequence, an iterative joint source-channel decoding (JSCD) technique is considered. To exploit the knowledge about the source correlations, we consider an iterative decoding between a sum-product (SP) decoder serially concatenated with BCJR decoder which is applied for every sensor as local iterations. Then, correlation between sensors' data is employed to update extrinsic information received from the SP-BCJR decoders of the different sensors during global iterations. We illustrate the performance of the joint decoder for different correlation setups and with different number of sensors. Simulation results, in terms of bit error rate show promising improvements compared with the separate decoding scheme where the correlation knowledge is not completely utilized in the decoder.