通信辐射源个体识别的自编码器构造方法  

Construction Method for Individual Communication Transmitter Identification Autoencoder

作　　者：黄健航 雷迎科 HUANG Jian-hang;LEI Ying-ke(School of Electronic Countermeasure,National University of Defense Technology,Hefei 230037,China)

机构地区：[1]国防科技大学电子对抗学院,合肥230037

出　　处：《火力与指挥控制》2018年第11期108-112,共5页Fire Control & Command Control

基　　金：国家自然科学基金(61272333);国防科技重点实验室基金(9140C130502140C13068);军队预研基金资助项目(9140A33030114JB39470)

摘　　要：自编码器算法具有不依赖有标签样本进行有效特征学习的特点,适合应用于电子对抗中通信辐射源个体识别任务,但其主要局限在于其性能取决于结构参数的设计。为提高通信辐射源个体识别任务中自编码器性能,提出了一种通信辐射源个体识别的自编码器构造方法。提取通信辐射源信号的高阶积累量,根据调制信息估计自编码器结构参数,最后由验证实验筛选性能满足阈值的结构参数,存入参数信息库。验证实验中,在实际采集的调频通信电台数据集上最高达86.8%的准确率,证明了所构造自编码器的有效性。Autoencoders can learn feature from input efficiently without dependence on large number of labeled samples,which is suited for individual communication transmitter identification in the area of electronic countermeasure.But it has a fatal defect that its performance is greatly influenced by its structure.A construction method is proposed to improve the performance of autoencoders for individual communication transmitter identification.Firstly,high-order cumulants of communication emitter signals are extracted and then used for estimating parameters of stacked auto-encoders.Finally,a set of network parameters meeting the performance requirement are selected by confirmatory experiment and saved in a parameter information library.The performance of network constructed by the method proposed reached the accuracy of 86.8%in experiments,in which the data is collected from frequency modulation radios in actual communication environment.As a result,the efficiency of the method proposed has been identified.

关 键 词：通信辐射源个体识别 自编码器 参数估计 

分 类 号：TN911.7[电子电信—通信与信息系统]