基于RBF神经网络的海杂波建模

从相空间重构理论出发，构造了一个RBF神经网络预测器来重构海杂波的内在动力学，并且利用这个确定性的模型时海杂波的演变进行预测。为验证模型的推广性能，采用了多步预测。对雷达采集的实际海杂波数据的实验结果表明，这个确定性的模型可以很好地追踪海杂波的演变。文中还分析了该RBF神经网络预测器在不同高斯白噪声条件下的预测性能，得出了其预测误差与杂噪比（CNR）的关系。

The Modeling of Sea Clutter Based on RBF Neural Network

With the guidance of the theory of phase space reconstruction, a RBF neural network predictor was designed to reconstruct the underlying dynamics of sea clutter, and the evolution of real sea clutter data collected by radar was predicted using the deterministic model. The recursive prediction was picked to demonstrate the generalization ability of the network model. The result shows that this deterministic model so obtained is capable of predicting the evolution of sea clutter. Furthermore, simulations using the RBF neural network were also performed with increasing white Gaussian noise levels, and the relation between the predictive error and the clutter-to-noise ratio (CNR) was educed.