纠双错神经网络非线性码的研究

Study of Double Error Correcting Nonlinear Code with Neural Network

本文献在文［１］、［２］的基础上，设计了一种离散型Ｈｏｐｆｉｅｌｄ自反馈神经网络，其权值矩阵仍具有较强的规律性。将这种网络应用于编码技术，得到了纠双错的非线性（ＤＥＣ－ＴＥＤ）码，同时对非线性码的旋转平移性进行了研究。这种网络原理简单，译码复杂性低，所得码具有一定的规律性。

In this paper, a new discrete Hopfield neural network with self feedback connection is designed, in which weight matrix has a better structure. By using this kind of neural network, double error correcting and three error detecting (DEC TED) codes are obtained. Furthermore, rotation and translation with the new codes are studied. Clearly, the neural network is simple in principle and has lower complexity in decoding. The code obtained has some regularity to be used.