基于复杂度特征的通信辐射源个体识别

发布时间：2018-03-21 08:17

本文选题：通信辐射源　切入点：调制识别　出处：《哈尔滨工程大学》2014年博士论文　论文类型：学位论文

【摘要】：当今的时代,是无线通信技术发展极其迅猛的时代,随着高新技术的快速发展以及战争形态的日益变化,信息战势必发展为未来战争的主要形态,而通信辐射源个体识别技术是信息对抗领域的关键技术之一。通信辐射源个体识别算法主要采用的是模式识别方法。模式识别方法的一般步骤是,首先对信号进行预处理,包括对信号进行一定的去噪处理或进行某种变换,然后对预处理后的信号进行分析,提取可以代表辐射源个体特征的参量保存到数据库,作为信号的特征参量。若截获的信号特征能与数据库中的信号特征匹配,则认为与该信号是同一辐射源发出,从而达到识别辐射源个体的目的。随着通信电磁环境的日益复杂,以及通信信号的样式的逐渐增多,如何在较低信噪比下有效的提取辐射源的个体特征,是各国学者关注的热点问题。针对辐射源个体识别中如何在低信噪比下有效的提取辐射源个体特征这一问题,论文提出了几种新的特征提取算法,并设计了灰色关联分类器,对提取到的特征进行分类识别,具体内容如下:由于各个特征提取算法的性能都要用最终对辐射源个体的识别效果来验证,即需要利用分类器对提取到的信号特征进行分类识别,因此,先对分类器的设计算法进行了介绍,以便于在后文中对分类器的使用。灰色关联理论主要是通过计算两个不同离散序列的关联度进而判断序列的相似关联程度。相对于神经网络分类器而言,其实时性识别能力强,但自适应能力较差,针对这一问题,首先,提出了改进的灰色关联算法,通过自适应的对特征序列中各个特征的重要程度的选择,来提高算法的自适应能力;又针对低信噪比条件下,提取到的信号特征往往呈现区间分布的这一特性,提出了改进自适应区间的灰色关联算法,仿真结果表明,该算法能够实现低信噪比下,对提取到的交叠信号特征进行分类的目的。其次,针对辐射源个体识别中特征提取这一模块,提出了基于熵云特征Holder云特征的二次特征提取算法。该算法通过计算低信噪比下不稳定的熵特征和Holder系数特征的分布特性,即提取信号第一次特征提取到的特征分布的均值、熵、超熵这3个云模型的数字特征,进一步对信号的离散分布特征进行特征提取,通过二次特征提取,更为精确的刻画了信号在低信噪比下的特征分布,再利用自适应区间灰色关联分类器对提取到的三维云特征进行分类,实现了低信噪比下的个体识别。再次,提出了基于改进分形盒维数的辐射源个体特征提取算法,对分形理论中的一维盒维数的基本算法进行了改进,通过对盒维数拟合曲线的每一点值进行求导,组成待识别信号的盒维数特征向量,更精细的对信号的盒维数特征进行了刻画,相对于传统的一维盒维数特征,具有更好的识别效果。最后,提出了基于多重分形维数的辐射源个体细微特征提取算法,对不同噪声环境下的通信电台个体信号或是携带不同电台内部细小噪声的电台信号进行多重分形维数的特征提取,通过提取离散信号不同重构空间下的微小特征,进而实现对细微特征进行识别的目的。
[Abstract]:In today's era, is the rapid development of wireless communication technology is the era, with the rapid development of high technology and changing the form of war, information warfare will develop as a main form of future wars, and communication emitter identification technology is one of the key technology in the field of information warfare. Communication emitter identification algorithm is mainly used is the pattern recognition method. The general steps of pattern recognition methods, the signal pretreatment, including signal denoising or some transformation, and then the signal preprocessing after analysis, extraction parameters can represent the individual characteristics of the radiation source is saved to the database, as if the characteristic parameters of signal. The signal characteristics can match the intercepted signal feature and the database, and that the signal is the same radiation emitted by a source, so as to achieve the recognition of individual radiation source The purpose of communication. With the increasingly complex electromagnetic environment, and gradually increase the style of communication signals, the individual feature extraction of radiation source in low SNR effectively, is a hot topic for many scholars in different countries. In order to solve the problem of feature extraction of radiation source in low SNR effective emitter the recognition of individual papers, this paper puts forward some new feature extraction algorithm, and designed a grey correlation classifier, the extracted features to the classification, the specific contents are as follows: the performance of various feature extraction algorithms are used for final emitter recognition results to verify that signal characteristics using the classifier to extract the classification, therefore, the design of classifier algorithm was introduced to facilitate the use of classifiers in this article. The grey relational theory is mainly through the calculation of the two is not Correlation with discrete sequence similarity degree and judge sequence. Compared to the neural network classifier, strong recognition ability in fact, but poor adaptability, in order to solve this problem, firstly, puts forward the grey correlation algorithm, the important degree of each feature in the sequence of adaptive selection to improve the adaptability of the algorithm; and according to the condition of low SNR, signal feature extraction to tend to the characteristics of interval distribution, puts forward the grey correlation algorithm of adaptive interval, the simulation results show that this algorithm can achieve low SNR, the extraction of overlapping signal features for classification purposes secondly, according to the characteristics of emitter identification in the extraction module, put forward the two feature extraction algorithm entropy cloud feature Holder cloud based on features. By calculating the low SNR Than the distribution characteristics of entropy feature and Holder coefficient characteristics under unstable, that is the first time to extract the signal feature extraction feature distribution mean, entropy, entropy of the 3 super digital characteristics of cloud model, further discrete distribution of signal feature extraction, extraction of syndrome through two more accurately portray the special. The signal in low SNR distribution, 3D cloud feature by adaptive interval grey correlation classifier for extraction to classify, to realize the low signal-to-noise ratio of individual identification. Thirdly, put forward improved emitter feature extraction algorithm based on fractal box dimension, the basic algorithm of one-dimensional fractal box dimension the theory was improved, the derivation through each point of the fitting curve of the box dimension, box dimension feature vector for recognition of the signal. The signal box dimension feature of more precise moment Painting, relative to the traditional one-dimensional box dimension feature, has better recognition effect. Finally, we propose an extraction algorithm of radiation source multi fractal dimension of individual subtle features based on individual communication transmitter signals in different noise environments or feature extraction of radio signals with different radio noise of multiple internal small fractal dimension, small by extracting the discrete signal characteristics of different reconstruction space, so as to realize the purpose of identification of subtle features.

【学位授予单位】：哈尔滨工程大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN911.7

【相似文献】