In practice, LDPC codes are decoded using message passing methods. Thesemethods offer good performance but tend to converge slowly and sometimes failto converge and to decode the desired codewords correctly. Recently,tree-reweighted message passing methods have been modified to improve theconvergence speed at little or no additional complexity cost. This paperextends this line of work and proposes a new class of locally optimizedreweighting strategies, which are suitable for both regular and irregular LDPCcodes. The proposed decoding algorithm first splits the factor graph intosubgraphs and subsequently performs a local optimization of reweightingparameters. Simulations show that the proposed decoding algorithm significantlyoutperforms the standard message passing and existing reweighting techniques.
展开▼
