Adaptive Normalized Min-Sum Decoding Algorithm For LDPC Codes in Flash Memory

For globally coupled low density parity check (GCLDPC) codes, the normalized Min-Sum (NMS) algorithm is a widely-used decoding algorithm but suffers from the degraded performance and the slow convergence speed. To address this issue, in this paper, by analyzing the relation between the bit error ratio (BER) and the degree of polarization of log-likelihood ratio (LLR) information, we propose an adaptive NMS decoding algorithm. In particular, the normalized factor is dynamically adjusted based on the average LLR information in the initial and the current iterations. Besides, the two-level decoding scheme is considered. Simulation results show that, in the additive white Gaussian noise (AWGN) channels and the flash memory channels, as compared with the NMS algorithm, our algorithm significantly improves the decoding performance and the convergence speed.