基于杂散特征的辐射源个体识别研究

发布时间：2018-10-25 19:46

【摘要】：作为通信对抗领域的关键技术,辐射源个体识别研究近年在国内外引起广泛关注。随着通信辐射源设备的组成日益复杂化,如何截获并分析出通信信号中所反映出的辐射源个体特征信息并将其有效分类,成为个体识别研究的热点内容。国内现有的研究成果中,辐射源个体特征的识别主要针对不同类别的设备进行研究,且特征分类需要依赖于充足的样本数量,而国外的研究成果集中于暂态信号的研究。在总结国内外相关研究的基础上,本文以同类辐射源个体的杂散特征作为研究对象,对辐射源杂散特征的提取方法和分类识别进行了研究。特征提取方面,本文对辐射源的稳态信号的边际谱杂散特征进行了提取。利用基于经验模态分解(EMD)和小波包重构的边际谱特征提取方法,得到边际谱的能量重心和信息熵。但在原有特征提取方法中,EMD方法存在模态混叠问题,且小波包重构标准需凭借经验选择。针对以上问题,提出一种改进方案,首先对EMD的模态混叠成因进行分析,并改进小波包重构的标准和EMD的输入,在求取信号边际谱时,采用叠加时频谱的方法,降低了模态混叠的影响,改善了杂散特征分布的聚集性差及掺杂程度严重的问题,特征分离程度得到一定的提高。分类识别方面,设置了不同数量样本条件下的两组分类实验,利用K近邻算法、神经网络分类器及支撑向量机(SVM)分类器对不同调制样式下的通信信号进行了分类性能比较,并从中选择出性能最优的SVM分类器对特征提取方法前后的分类效果进行比较,仿真结果表明,改进的特征提取方法使辐射源信号特征的分类性能得到一定程度的提高,在不同调制类型下具有较好的适用性。
[Abstract]:As a key technology in communication countermeasure field, the research on individual identification of emitter has attracted wide attention at home and abroad in recent years. With the increasing complexity of communication emitter equipment, how to intercept and analyze the individual characteristic information of emitter reflected in communication signal and classify it effectively has become a hot topic in individual identification. Among the existing research results in China, the identification of individual characteristics of emitter is mainly focused on different types of equipment, and the classification of features depends on the number of sufficient samples, while the research results abroad focus on transient signals. On the basis of summing up the relevant research at home and abroad, this paper takes the stray characteristics of the same radiation sources as the research object, and studies the extraction method and classification recognition of the stray characteristics of the same radiation sources. In the aspect of feature extraction, the marginal spectrum stray feature of steady state signal of emitter is extracted in this paper. Based on empirical mode decomposition (EMD) and wavelet packet reconstruction, the energy center of gravity and information entropy of marginal spectrum are obtained. However, in the original feature extraction method, the EMD method has the problem of modal aliasing, and the wavelet packet reconstruction criteria need to be selected by experience. In order to solve the above problems, an improved scheme is proposed. Firstly, the causes of modal aliasing in EMD are analyzed, and the standard of wavelet packet reconstruction and the input of EMD are improved. The method of superposition time spectrum is used to calculate the marginal spectrum of the signal. The influence of modal aliasing is reduced, the aggregation of stray feature distribution is improved, and the problem of serious doping degree is improved, and the degree of feature separation is improved to a certain extent. In the aspect of classification and recognition, two groups of classification experiments with different number of samples are carried out. Using K-nearest neighbor algorithm, neural network classifier and (SVM) classifier of support vector machine, the classification performance of communication signals with different modulation patterns is compared. The SVM classifier with the best performance is selected to compare the classification effect before and after the feature extraction method. The simulation results show that the improved feature extraction method can improve the classification performance of the emitter signal to a certain extent. It is suitable for different modulation types.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN975

【相似文献】