Adaptive normalized min-sum algorithm for LDPC decoding

Abstract

This paper proposes an adaptive normalized minsum algorithm for the decoding of low-density parity check (LDPC) codes, which utilizes an adaptive normalization factor to improve the accuracy of the soft information transferred during the iterative decoding process, and provides superior performance accordingly. This adaptive normalization factor can be adjusted dynamically and adaptively at each decoding iteration according to a look-up table obtained via training and simulation. Its implementation facility and independence from the channel characteristics make the proposed adaptive normalized min-sum algorithm expect a wide application. Simulation results show that the proposed algorithm can achieve performance much closer to the sum-product algorithm and a coding gain of around 0.2dB compared to the conventional normalized min-sum algorithm for DVB-S2's rate 2/5 and 3/5 LDPC codes over the additive white Gaussian noise (AWGN) channel.

NOLD: A neural-network optimized low-resolution decoder for LDPC codes

“…When it reaches the maximum iteration number I, the variable-to-check processing at iteration I can be represented by (13). As shown in ( 17) and ( 18), within one iteration, the outgoing messages from check nodes or variable nodes are scaled and quantized equally for each node.…”

“…The typical way is to optimize the values of scaling factors by the adaptive strategies, which design an adaptive rule that computes the optimal scaling factors from the decoder's feedback at each iteration. The work of [13] utilized a pre-configured lookup table, wherein wherein the proportion of unsatisfied paritycheck equations adaptively determines the scaling factors. It has been shown in [14] that the scaling factors increase exponentially with the iterations and the final values are equal to 1.…”

Adaptive Quantized and Normalized MSA Based on Modified MET-DE and Its Application for 5G-NR LDPC Codes

“…To improve the threshold performance of MSA, offset MSA (OMSA) [16] and normalized MSA (NMSA) [17] are proposed, and the latter retains the robustness of MSA. However, conventional NMSA uses fixed normalization factor, leading to a performance tradeoff between the decoding threshold and the error floor [18]. Many efforts have been made to enhance the decoding performance of OMSA and NMSA.…”

“…Fan et al propose an improved NMSA called adaptive NMSA (ANMSA) [18], which utilizes a preconfigured look-up table (LUT) between the normalization factor α and the unsatisfied check nodes rate p, also referred to as check-ratio, and uses Monte Carlo simulation to guide the LUT design. or check nodes of irregular LDPC codes [27].…”