基于 VMD 和熵特征的雷达辐射源信号识别-《计算机技术与发展》

文章信息/Info

Author(s):: YANG Jie; YUE Mei-jun; ZENG Yao-ping; School of Communication and Information Engineering,Xi’an University of Posts and Telecommunications,Xi’an 710121,China

Keywords:: variational mode decomposition; center frequency; feature fusion; particle swarm support vector machine; entropy feature

摘要:: 将雷达信号的近似熵(ApEn) 和范数熵( NoEn)提取出构成特征向量,用粒子群优化的支持向量机进行分类识别,得到结果发现对于相似的特征向量识别正确率较低。为了提高雷达辐射源个体的识别正确率,提出了一种基于变分模态分解(variational mode decomposition,VMD) 和熵特征相结合的多维特征雷达辐射源信号识别方法。首先利用 VMD 算法对各雷达信号进行分解,得到两个本征模态组合函数,因为不同的雷达信号分解成各个模态的中心频率也是不同的,然后组合中心频率特征与近似熵、范数熵特征进行特征融合构成 4 维特征向量,最后使用粒子群支持向量机( particle swarmoptim-ization support vector machine,PSOSVM) 对辐射源信号进行识别。实验结果表明,经过特征融合构成的特征向量展示了更好的识别效果。集聚多种特征的识别优势来提升雷达辐射源信号识别准确率,相比于原来单一熵特征结合识别方法在分类效果上更具有优势。

Abstract:: The approximate entropy ( ApEn) and norm entropy ( NoEn) of the radar signal are extracted to form the feature vector,and the support vector machine optimized by particle swarm is used for classification and recognition. The result is that the recognition rate of similar feature vectors is low. In order to improve the recognition accuracy of individual radar emitters,a multi-dimensional feature radar emitter signal recognition method based on the combination of variational mode decomposition ( VMD) and entropy features is proposed.First,the VMD is used to decompose each radar signal to obtain two eigenmode combination functions,because the center frequency of different radar signals decomposed into each mode is also different. Then the center frequency feature is fused with approximate entropy and norm entropy feature to form a 4 - dimensional feature vector. Finally, particle swarm optimization support vector machine( PSOSVM) is used to identify the radiation source signal. The experiment shows that the feature vector composed of feature fusion has a better recognition effect. Integrating the recognition advantages of multiple features to improve the accuracy of radar emitter signal recognition is more advantageous than the traditional single entropy feature combined recognition method.