雷达辐射源信号聚类分选算法

Deinterleaving technology of radar emitter signals based on clustering

基于脉内特征参数的聚类是实现雷达辐射源信号分选的一种重要途径。本文在小波域滤波算法的基础上提出一种新的雷达辐射源信号脉内特征提取和聚类分选方法,将小波变换后的低频逼近小波系数的能量分布熵与经过尺度相关去噪计算后反映信号边缘的高频细节小波系数能量分布熵作为分选的脉内特征向量,并引入灰关联测度来衡量脉内特征样本之间的相似程度,实现雷达辐射源信号的聚类分选。通过对10种不同雷达辐射源信号的仿真实验分析表明:提取的样本特征在低信噪比(0dB)下具有很好的抗噪性。同时与FCM聚类相比,基于灰关联测度的层次聚类算法具有更高的分类正确率。

Clustering algorithm based on intra-pulse features is an important approach to realize deinterleaving of radar emitter signals.In this paper,a novel algorithm based on wavelet transform domain filtering is proposed to deinterleave radar emitter signals.The energy entropy from approximation coefficients of wavelet transform and the other energy entropy from inner-scale correlations denoise of detail coefficients are used as deinterleaving feature vector.And grey relational analysis,introduced to measure the degree of similarity between feature vector sets,is used to sort different radar emitter signals.Computer simulation results and clustering of 10 different radar signals show that the feature vectors have good performance of noise-resistance when SNR is 0dB.Moreover,compared with FCM clustering algorithm,the hierarchical clustering algorithm based on grey relational analysis has higher accuracy of clustering.