Low-Complexity Belief-Propagation Decoding via Dynamic Silent-Variable-Node-Free Scheduling

Partial IDS decoding based on the base graph of protograph LDPC codes

“…In this section, the decoding performance of the PABRBP algorithms is tested for a large number of simulations. To verify the outstanding decoding performance of the proposed algorithm, many of the most representative algorithms, such as Serial Belief Propagation (CSBP) [10], Log-Likelihood Ratio Belief Propagation (LLR BP) [1], Residual Belief Propagation (RBP) [11], Node-Wise Residual Belief Propagation (NW-RBP) [11], Lazy Queue Residual Decoding Algorithm (LQRD) [13], Dynamic Silent Variable Node Free Scheduling (D-SVNFS) [17], and Residual-decaying-based Residual Belief Propagation (RD-RBP) [18]. The Decoding Algorithm Based on Random Select of Check Nodes with An Adjustable Update Range (RSCAR) [21] and The Decoding Algorithm Based on Random Select of Check Nodes with A Predefined Update Range (RSPUR) [21] are also used for comparison.…”

“…The error‐correction performance of the PABRBP algorithm is significantly better than that of any other IDS algorithm. Figure 4 shows that when BER is $$1.0\times 10^{-4}$$, the PABRBP algorithm can obtain an almost 0.3 dB gain compared with the D‐SVNFS [17] algorithm. Figure 5 shows that when FER is $$1.0\times 10^{-3}$$, the PABRBP algorithm can obtain a gain of almost 0.3 dB compared with the RSCAR algorithm.…”

“…A single iteration of the RBP algorithm involves $$E(d_v-1)$$ updates for V2C messages and E updates for C2V messages. The NW‐RBP [11], LQRD [13], Q‐RBP [16], D‐SVNFS [17], and RD‐RBP [18] algorithms, derived from the RBP algorithm, share identical updating complexity. In IDS algorithms, decoding performance is mainly affected by comparative complexity.…”

“…Currently, many IDS algorithms [11][12][13][14][15][16][17][18][19][20] have shown promising decoding performance, but the resource allocation and greedy problems have not been adequately solved. These previous IDS algorithms always search and update the edge with the maximum message-residual from all the edges, which signifies that all the edges will be searched and calculated for each iteration.…”

“…By reasonably allocating computational resources, the Q‐RBP algorithm achieves more attractive decoding performance than the above algorithms. Afterwards, the dynamic silent‐variable‐node‐free scheduling (D‐SVNFS) algorithm arranges the update order according to the reliability of variable nodes [17], and then, the intrinsic messages of each variable node are propagated during the process of decoding, which substantially enhances the error‐correction performance. To improve the convergence performance, the residual‐decaying‐based RBP (RD‐RBP) algorithm has been proposed by weighting the residual of the edges with a factor less than 1 [18].…”