基于变量节点串行消息传递的LDPC码译码研究

A Serial Decoding Schedule for LDPC Based on a Serial Update of Bit Nodes' Messages

针对标准LDPC码译码中洪水消息传递机制的不足,提出以串行机制进行消息传递,按照变量节点的顺序进行消息处理和传递,对每个变量节点同时接收校验消息和发送变量消息。该方法使更新的消息能够很快进入当前迭代计算,改善了LDPC迭代译码的收敛性能。通过对几种常用译码算法的仿真比较,验证了在复杂度不增加的情况下,该方法性能优于其它几种最大后验概率准则的译码方法,且算法收敛快,是一种能较好兼顾性能与实现复杂度的译码方法。

Contraposing the shortcoming of ＂flooding schedule＂ in the standard LDPC （ low-density parity-eheek ） decoding algorithm, this paper presents a serial decoding schedule, which shuffles the massage passing schedule of standard LDPC decoding algorithm. The new schedule processes and propagates messages according to a serial update order of bit nodes. Each bit node receives cheek messages and sends bit messages in the same iteration so that the updated messages could join iterafive process immediately. As a result, the convergence property is improved. Simulation results show that, under the same decoding complexity, the proposed version is optimal in the sense of MAP decoding and converges faster than the flooding schedule. Meanwhile, it keeps the appropriate decoding complexity, so it is a good method that can offer better trade-offs between performance and complexity.