电台个体识别研究

发布时间：2019-01-18 14:46

【摘要】：通信电台个体识别技术是通过某种方法来获取并检测出表征通信电台的个体“指纹”特征,从而实现个体的识别。本文首先从通信电台的硬件特性出发研究了“指纹”特征的产生机理。并结合特征提取、选取及分类方法的国内外发展趋势,对全文的研究内容安排及论文结构做了探讨。基于信号分析技术的发展更迭顺序,本文研究了从傅里叶分析,小波分析到近年来兴起的Shearlet分析的信号时频分析理论,为本文研究通信电台个体识别技术奠定理论基础。基于经典的模糊函数分析方法,本文利用拥有优异时频分析性能的小波分析方法和新兴的可以表征各向异性的Shearlet分析方法对其进行了一定的改进,并将改进后的方法运用于电台个体识别。不同于经典的模糊函数方法,改进方法在对信号的自相关处理之后,利用了三次分段B样条小波、db小波以及Shearlet分析取代原本的傅里叶分析,以期做到更好的特征捕获。并且针对MSK调制以及PSK调制的仿真信号进行试验分析,仿真结果表明,改进方法的效果受调制方式的影响较小,模糊函数方法的效果受调制方式的影响较大。在MSK调制下,相对于传统的模糊函数方法,db4小波分析方法、B样条小波分析方法以及Shearlet分析方法都具有更好的抗噪性能,其中三次B样条小波分析方法效果更优也更为稳定。最后,本文研究了信号特征的降维与分类。为了避免“维数灾难”,在信号分类的时候,需要将特征降维到一定程度,再运用一个好的分类器进行分类,这样将对分类效果起着提升的作用。在研究了Fisher降维法、KNN分类器以及SVM分类器之后,通过实验仿真去检验了不同分类器对于最终识别率的影响程度。
[Abstract]:The technology of individual identification of communication station is to obtain and detect the characteristic of individual "fingerprint" which represents the communication station by a certain method, so as to realize the recognition of individual. In this paper, the mechanism of fingerprint feature is studied based on the hardware characteristic of communication station. Combined with the development trend of feature extraction, selection and classification methods at home and abroad, the research content arrangement and the structure of the thesis are discussed. Based on the development sequence of signal analysis technology, this paper studies the theory of signal time-frequency analysis from Fourier analysis, wavelet analysis to Shearlet analysis, which lays a theoretical foundation for the research of individual identification technology in communication stations. Based on the classical fuzzy function analysis method, the wavelet analysis method with excellent time-frequency analysis performance and the new Shearlet analysis method, which can represent anisotropy, are used to improve it. The improved method is applied to individual recognition of radio station. Different from the classical fuzzy function method, the improved method uses cubic B-spline wavelet, db wavelet and Shearlet analysis to replace the original Fourier analysis in order to achieve better feature acquisition. The simulation results of MSK modulation and PSK modulation show that the effect of the improved method is less affected by the modulation method, and the effect of the ambiguity function method is greatly affected by the modulation method. Compared with the traditional fuzzy function method, db4 wavelet analysis, B-spline wavelet analysis and Shearlet analysis have better anti-noise performance under MSK modulation, and the cubic B-spline wavelet analysis method is more effective and stable. Finally, the dimension reduction and classification of signal features are studied. In order to avoid "dimensionality disaster", it is necessary to reduce the dimension of the feature to a certain extent and then use a good classifier to improve the classification effect. After studying the Fisher dimensionality reduction method, KNN classifier and SVM classifier, the influence of different classifiers on the final recognition rate is tested by experimental simulation.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN924

【参考文献】