Investigation On Graph-based Decoding Schemes For LDPC Codes

Low-density parity-check(LDPC)codes,which are a kind of error-correcting codes with excellent performance,have been widely used in a lot of communication standards.Iterative decoding algorithms processing the check nodes and variable nodes parallelly or in several groups are usually adopted by LDPC codes;however,they have high computational complexity.In order to solve the problem and enhance flexibility of designing decoding algorithms,an improved decoding scheme is proposed in this paper,according to the special structures of submatrices decomposed from the parity check matrices of LDPC codes.First,considering that Tanner graphs of LDPC codes can be decomposed arbitrarily,parity check matrices are decomposed and analyzed in this paper.Taking non-binary cycle codes for an example,a matrix decomposition method that is suitable for LDPC codes with parity check matrices having special structures is presented.Then,an improved decoding scheme is designed to achieve performance gains of LDPC codes.In this scheme,belief propagation(BP)algorithm or Log-MAP algorithm is used to decode submatrices with different structures,respectively.Furthermore,the performance of non-binary cycle codes employing the proposed algorithm is evaluated.Simulation results show that the proposed algorithm performs better than BP algorithm under the small iterations.Finally,in comparison with BP algorithm,the decoding complexity is also analyzed in this paper.