基于特征降维的卫星通信辐射源个体识别方法

摘要

A novel radiometric identification algorithm based on reconstructive dimensional reduction analysis is proposed for the applications of satellite communication security.The proposed algorithm searches the subspace which guarantees the minimum distances of intra-class feature vectors and the maximum distances of inter-class feature vectors in a supervised fashion after the extraction of the high-dimension feature vectors from all satellite devices.After the reduction of the feature vectors,the support vector machine can be trained and be used to predict the class label of an unknown device.The discriminative fingerprint features of satellite devices extracted with the proposed algorithm are linear combinations of high-dimension vectors,which reserve all the minute differences of different satellite devices.Experiments on actual data sets show the effectiveness of this algorithm for tasks of satellite devices identification.%针对卫星通信系统安全检测问题,提出了一种基于重构降维分析的卫星辐射源个体识别方法.该方法通过提取所有卫星终端设备训练数据的高维特征向量,然后以监督方式寻找使得降维特征向量离同类最近、离异类边缘点最远的子空间,并利用降维特征向量训练分类器,最后利用分类器判决未知信号的类别.提出的卫星辐射源个体识别方法所提取的辐射源特征基于高维特征向量的线性组合,保留了不同发射机的差异信息,具备很强的分类辨别能力.实际采集的数据测试结果表明,该方法可有效识别不同辐射源个体.