基于学习矢量量化(LVQ)神经网络的雷达体制识别

由于现代战争中雷达体制的多样化、复杂化及其综合应用使得雷达体制识别中要处理大量复杂的高维数据,学习矢量量化(LVQ)神经网络不仅能处理有监督分类,而且相对于其他神经网络能以较小的计算量处理大量输入数据,所以采用LVQ对雷达体制进行识别,同时针对LVQ学习速率的变化可能引起学习算法不稳定,采用修正的学习速率算法.在简要介绍雷达体制和LVQ的基础上构造了LVQ神经网络对雷达体制进行分类.通过与径向基神经网络(RBFN)识别算法的仿真对比,证实了方法的有效性.

Radar-Type Recognition based on LVQ Neural Network

Because of the diversification,complication and intergrated application of radar types in morden warfare,recognizing radar types needs treating many high-volume data.The learning vector quantization(LVQ) can treat many input data with small computational burdens compared with other neural network.So we use the LVQ network to classify the radar types.This paper adopted a modificatory learning rate method to deal with the instability bringing from the change of LVQ's learning rate.Compared with radial basis function network(RBFN),which has been adopted as a classification method,the proposed method has shown excellent classification results.