基于GMM和神经网络的辐射源识别方法

针对基于截获雷达脉冲特征参数的辐射源识别问题,通过建立一个高斯混合模型(GMM),采用最大化期望(EM)方法对模型参数进行训练,构建了一个输入为截获雷达脉冲特征参数,输出为雷达辐射源类型的分类器。同时,为实现对分类识别性能对比,进一步提出基于神经网络方法构建雷达辐射源类型分类器。仿真试验结果表明,基于GMM和神经网络构建的两种分类器均能实现对雷达辐射源的在线识别,且当用于训练的样本比例不低于10%时,均能获得90%以上的分类正确率。

The Methods Based on the GMM and Neural Network for Recognition of Emitters

Considering the recognition of emitters based on the parameters of interception radar pulse,a Gaussian mixture model（GMM）is built and trained by the expectation maximization（EM）method,so a clas-sifier is constructed whose input is the interception radar pulses and whose output is radar emitter types. Then,another classifier based on the neural network method is also compared with the proposed GMM-based method.The results of extensive simulations demonstrate that the proposed classifiers based on the GMM and neural network can achieve the on-line recognition of radar emitters,and the accuracy is more than 90%when the training sample ratio is not less than 10%.