Abstract:

To improve the performance of sparse check matrix reconstruction for low density parity check（LDPC）codes with high bit error rate and small number of received codewords, an algorithm combining error-containing codeword zeroing and layered belief propagation（LBP）decoding is proposed. This algorithm is based on random extraction and Gaussian elimination to obtain sparse check vectors, which are then used to verify the received codewords containing errors. Codewords that fail the verification are set to zero vectors, thereby increasing the number of sparse check vectors obtained in the early iterations of random extraction. When the number of sparse check vectors reaches a threshold, the LBP decoding method is used to correct the error bits, enhancing the reconstruction rate of the LDPC code’s sparse check matrix. Simulation results show that for the （648, 324） LDPC code under the IEEE 802.11n protocol, the proposed algorithm achieves a sparse check matrix reconstruction rate of 84% under conditions of high bit error rates and a small number of received codewords, which is an improvement of approximately 40% compared to existing algorithms. This effectively improves the fault tolerance of the resconstruction algorithm.

Key words: low density parity check（LDPC）, sparse check matrix, randomly selected, zeroing error codes, rebuild algorithm