摘要
本文提出了一种源于汉明类多层前向神经网络分组码译码器。它不需要改变网络结构和参数,根据输入可完成硬判决、软判决和最小距离译码。该网络不存在漫长的学习过程。计算机模拟表明,在加性高斯噪声下,使用该神经网络可以达到最大似然译码。
This paper presents a neural network decoder of linear block codes which originates from Hamming network. It can complete not only hard-decision decoding but also soft - decidion and minimum weighted distance decoding without changing the network' s structure and parameters. Moreover, the learning process is very fast and simple. The computer simulation shows that the decoder can achieve maximum likelihood decoding in the environmemts of additive white Gaussian noise.
关键词
线性分组码
神经网络
能量函数
最大似然译码
linear block code, neural network, energy function, maximum likelihood decoding
