Iterative joint decoding for sensor networks with binary CEO model

Abstract

An iterative joint decoding algorithm for data gathering wireless sensor networks is proposed in where the correlation between sensorspsila data is considered as a global code and iterative decoding is performed by concatenating the global decoder with the decoder of error correcting code applied to encode sensors observations. We apply this algorithm for sensor networks with binary CEO model where sensors observe different noisy versions of a single source, located away from sensors. This calls for employing more powerful error correcting codes, therefore we apply convolutional codes (Hamming codes and single parity check codes are applied in). We use the concept of iterative horizontal-vertical decoding for concatenated block codes to formulate the update rules for L-values for the considered binary CEO model. Our simulations confirm that the iterative joint decoding scheme substantially decreases the bit error rate compared with the separate decoding scheme, and reaches the minimum achievable distortion for channels with significantly higher noise levels.