Abstract | ||
---|---|---|
For the linear programming decoding (LPD) proposed by Feldman et al., the number of constraints increases exponentially with check degrees. By decomposing a high-degree check node into a number of degree-3 check nodes, the number of constraints grows linearly with check degrees. In this letter, we show that the size of the LPD can be reduced by decomposing a high-degree check node into a number of degree-4 check nodes. The LPD using the degree-4 decomposition leads to almost the same number of constraints as using the degree-3 decomposition, while the number of auxiliary variable nodes is less than half of the one using the degree-3 decomposition. Moreover, when decomposing a high degree check node into a number of check nodes with degree d, d>4, the number of constraints increases rapidly and the size of the LPD becomes larger than the degree-4 decomposition. It is demonstrated on an LDPC code and a BCH code that the decoding time of the degree-4 decomposition is the smallest among the different decomposition methods. |
Year | DOI | Venue |
---|---|---|
2013 | 10.1109/LCOMM.2012.122012.122396 | IEEE Communications Letters |
Keywords | Field | DocType |
Maximum likelihood decoding,Linear programming,Block codes,Iterative decoding,Complexity theory | Discrete mathematics,Low-density parity-check code,Linear programming decoding,BCH code,Linear programming,Decoding methods,Longitudinal redundancy check,Mathematics,Exponential growth,Decomposition | Journal |
Volume | Issue | ISSN |
17 | 2 | 1089-7798 |
Citations | PageRank | References |
0 | 0.34 | 0 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Xiaopeng Jiao | 1 | 38 | 9.90 |
Jianjun Mu | 2 | 41 | 10.63 |