基于改进的半监督AP聚类雷达信号分选

摘要

Radar signal sorting is a crucial aspect in modern high-tech warfare and information warfare,and is also an im-portant part of electronic warfare. In the perspective of time-frequency analysis,the signals of different modulation modes have different decomposition coefficients in some bands,and these differences allow the features extracted by the wavelet packet to recognize the modulation modes of signals. The cluster analysis of the statistical characteristics of signal wavelet packet energy entropy is performed by the improved semi-supervised affinity propagation clustering algorithm. Compared with other two algo-rithms,the improved semi-supervised affinity propagation algorithm is more effective and its accuracy is higher. In addition,the number of clustering is closer to the actual number.%雷达信号分选是现代高技术战争和将来信息化战争中至关重要的环节，是电子战的重要组成部分。从时频分析的角度出发，不同调制方式的信号会在某些频带内具有不同的分解系数，这些差异性使得小波包提取的特征能够进行信号调制方式的识别。提取信号小波包能量熵的统计特征，利用改进的半监督近邻传播聚类算法对特征进行聚类分析。与直接半监督近邻传播聚类算法和传统近邻传播算法比较，改进的半监督近邻传播聚类算法性能更优，准确率更高，而且聚类数更加接近实际聚类数。