论文部分内容阅读
给出的神经网络译码器是为长约束度（Ｋ≥１１）卷积码译码而设计的。Ｖｉｔｅｒｂｉ译码和序列译码是两种最大似然译码方法，虽然这两种技术能有效地提高误比特率性能，但它们都存在局限性。另外，只要在神经元和数字异或门单元之间建立局部连接，就能非常容易地直接用超大规模集成电路（ＶＬＳＩ）实现硬件。
The proposed neural network decoder is designed for long constraint (K≥11) convolutional code decoding. Viterbi decoding and sequence decoding are two kinds of maximum likelihood decoding methods. Although these two techniques can effectively improve the bit error rate performance, they all have limitations. In addition, hardware can be very easily implemented directly with very large scale integrated circuits (VLSI) as long as the local connections between neurons and digital XOR units are established.