论文部分内容阅读
在构造出分组码格图的基础上，利用一种基于前馈神经网络的多输入最小值选择网络实现分组码的软判决及硬判决译码．计算结果表明，前馈神经网络总能找到全局最优解，从而使该译码算法的性能同于最大似然译码．由于该前馈网络的计算时延非常短，且基于它的译码器与传统译码器相比硬件实现简单，从而使译码的复杂性降低，时延减小．
On the basis of constructing grid chart of block code, soft code decision and hard decision decoding of block code are realized by using a multi - input minimum value selection network based on feedforward neural network. The calculation results show that the feedforward neural network can always find the global optimal solution so that the performance of the decoding algorithm is the same as the maximum likelihood decoding. Due to the very short computation delay of the feedforward network and the simple hardware implementation of the decoder based on its decoder compared with the traditional decoder, the decoding complexity is reduced and the delay is reduced.