神经网络能量函数在卷积码译码算法中的应用  

The Use of Neural Network's Energy Function in theConvolution Decoding Method

作　　者：陈子勤[1] 郑义[1] 

机构地区：[1]汕头大学电子系,汕头515063

出　　处：《汕头大学学报（自然科学版）》2003年第4期39-43,共5页Journal of Shantou University：Natural Science Edition

摘　　要：文章分析了神经网络在信道编码领域的应用 .利用神经网络的退火算法 ,建立了新的基于 Hopfield神经网络的能量函数 ,借助卷积码的栅格 ( Trellis)图和新的能量函数提出了一种卷积码译码算法 ,仿真结果显示新的译码算法和最大似然 ( ML)译码算法的性能相似 ,该译码方案复杂度降低 。In this paper,we discussed the convolution decoding based on the Hopfield neural network.Using the simulated annealing arithmetic we predigested the energy function of the neural network.On the basis of the trellis diagram of the convolution codes and the new energy function of the neural network,a new neural network decoding method of the convolution decoding is presented.The simulation result shows that the network can always find global optimum.The performance of the decoding method approaches that of the ideal maxim likelihood decoding and it can greatly reduce the complexity and can be easily implanted by the hardware.

关 键 词：神经网络 能量函数 卷积码 译码算法 信道编码 栅格图 

分 类 号：TN911.22[电子电信—通信与信息系统]