Tree-Search Techniques for Joint Iterative Compressive Sensing and LDPC Decoding in Wireless Sensor Networks

Novel techniques are conceived for joint compressive sensing (CS) and low-density parity check (LDPC) coding in wireless sensor networks (WSNs), namely, a soft-input soft-output (SISO) tree search sphere decoding (SD) technique an SISO Hamming distance (HD)-based solution. Factor graphs utilized to describe the connectivity between signals sensors, as well with LDPC codes. In fusion center (FC), factor may be used iterative LDPC-CS order recover observed. However, CS decoder of FC suffers from high complexity if exhaustive Maximum A Posteriori (e-MAP) is employed, which considers all possible combinations source detected by each associated sensors. Hence, proposed SD HD schemes, only more likely tested reducing complexity. More specifically, first step find most combination signal values. Then, second step, continues set alternative hypotheses. This facilitates generation high-quality extrinsic information, iteratively exchanged decoder. By contrast, approach, obtains hypotheses within certain combination. Both our BLock Error Rate (BLER) results EXtrinsic Information Transfer (EXIT) charts show that approach performance full-search e-MAP at significantly reduced particular, we solution about 56 times complex than around 210 approach. Compared separate source-channel (SSCC) hard information benchmarker, schemes improve ${1}.{7} \text {dB}$ . Furthermore, allow iterations inside eliminate error floors obtain further notation="LaTeX">${2}.{45}$ -dB gain.