回声状态网络混沌跳频码预测方法

Chaotic Frequency Hopping Code Prediction Method Based on Echo State Network

针对现有跳频码预测方法存在的缺乏记忆能力、运算量大、训练过程复杂等问题,提出了基于回声状态网络的混沌跳频码预测方法。该方法在跳频码相空间重构的基础上,利用回声状态网络内部动态储备池的循环记忆功能,通过调整各权值矩阵的数值大小达到记忆数据的目的,解决了跳频码预测的问题。仿真实验表明该方法对Logistic-Kent映射、Lorenz系统和Mackey-Glass系统三种混沌跳频码都有较好的预测效果,并与其他方法的实验结果进行了比较,证明回声状态网络在混沌跳频码预测方面的可行性及优越性。

As for the problem of frequency hopping prediction such as the incapability of memorization, vast computation and complex training procedure, a new method for chaotic frequency hopping codes prediction based on the echo state network was proposed. The method, which was under the premise of the phase space reconstruction, solved the problem by taking the advantage of the cyclic memory function of dynamic reservoir and adjusting the numerical size of each weight matrix to achieve the purpose of memorizing data. The simulation experiments showed that the method achieved great prediction performance for the three chaotic frequency hopping codes generated by Logistic-Kent mapping, Lorenz and Mackey-Glass system respectively.