【作者】 徐书华；

【导师】 黄本雄；

【作者基本信息】 华中科技大学 ， 信息与通信工程， 2007， 博士

【摘要】 通信辐射源个体识别是近来通信对抗领域一个重要的研究课题,它主要根据各通信设备硬件差异在发射信号上所表现出来的区别于其它个体的特征,判别信号来自哪部通信设备,实现设备追踪,进而有针对性的对敌重要通信装备及其载体进行监视、电子干扰或者军事打击。不同于传统通信信号侦察中的调制模式识别研究,通信辐射源个体识别主要研究体现同类辐射源之间个体差异的信号指纹的分析提取技术。目前,国内外相关研究主要针对暂态通信信号中的开机信号、利用特征提取的分析方法实现辐射源个体识别。然而,利用暂态信号特征进行辐射源识别面临在非协作通信条件下对信号的捕捉、因暂态信号与噪声相似性带来的特征提取难度等诸多挑战;而且,这些方法大多通过仿真手段在较高信噪比和样本充足的条件下研究不同型号的通信辐射源识别问题。实际上,通信对抗环境中截获信号的信噪比往往较低,信号持续时间较短、样本数据不足,现有方法直接用于外场同类通信辐射源个体辨识时往往识别率较低。基于此,本文针对相同型号和工作模式不同通信辐射源的个体识别问题,对稳态通信信号的个体细微特征,即信号指纹的分析提取技术难点展开研究。本文研究了信号指纹的基本理论,从其产生机理出发,研究了通信设备在稳定工作状态下频率特性、调制参数特性以及杂散特性的个体差异,并从时域、频域、时频域和高阶谱等不同角度采用多种通信信号处理手段探索信号指纹特征的提取方法,建立了通信辐射源个体识别基本框架。文中提出了一系列具有理论及实用价值的算法,并全部通过外场实测电台数据验证了所提算法的优良性能。本文所做的工作主要包括以下几个方面:(1)研究了基于载频和码速率的信号指纹提取技术。给出了一种改进的相位拟合载频估计方法;并给出了一种基于STFT时频能量分布和小波变换的码速率估计方法,解决了非协作通信条件下调制信号的码速率估计精度问题。实验表明:载频和码速率估计结果可以作为信号指纹特征之一,并配合其它特征对电台个体进行识别。(2)研究了基于辐射源个体杂散输出成分的信号指纹提取技术。首先提出了一种基于正交分量重构的包络提取算法,并利用分形维数和Lempel-Ziv等复杂度特征提取个体信号的包络寄生调制特征;然后给出了利用个体信号Hilbert边缘谱对称性特征和HHT时频分布灰度图像抽取特征的细微差异进行辐射源识别的方法。实验表明:较低信噪比下信号个体的包络寄生调制特征和时频分布灰度图像抽取特征作为信号指纹参量对电台个体具有较好的聚类性能,而受载频估计精度影响,谱对称性特征对电台个体的可分离性不高。(3)研究了基于高阶谱的信号指纹提取方法。提出了一种矩形积分双谱(SIB)信号指纹特征分析方法,并利用局部线性嵌入(LLE)流形约简方法对高维SIB特征进行降维分析。实验表明:SIB较其它局部双谱特征用于个体信号的指纹分析具有一定的优越性,而且SIB约简特征具有较好的聚类性能和抗噪能力。(4)针对辐射源个体识别实际应用环境,设计了一种基于核距离测度的多类别SVM分类器(KDM-SVM)和一种基于D-S证据理论的组合分类器,并采用上述分类器,以不同角度提取的信号指纹特征参量集为分类依据进行了通信电台个体识别实验。实验结果表明:利用稳定工作状态下提取的信号指纹特征集进行电台个体识别可以获得良好的识别效果。本文的研究工作在一定程度上解决了在小样本、较低信噪比和变化信噪比条件下对实际同型号通信电台的个体识别问题,具有一定的理论意义和实际应用前景。更多还原

【Abstract】 As a very important subject in the filed of military communications confrontation, in-dividual transmitter identification technique is able to identify individual equipment using the subtle features of individual transmitted signal resulting from differences in hardware, and then the transmitter tracking, targeted surveillance, electronic jamming or military at-tack to the enemy’s electronic equipment and their carriers can be enacted. Different from modulation mode identification, individual transmitter identification mainly researches the extraction of individual subtle features between transmitters with the same model.Currently, the transient feature extraction using the boot signal is mainly researched for individual transmitter identification. However, transmitter identification based on tran-sient features faces some challenges, such as transient signal capture in non-cooperative communication environment and difficulties in feature extraction caused by similarity between transient signal and noise. Also, most methods work in higher signal-to-noise with sufficient samples, and mainly solve the problem of identifying transmitters with dif-ferent models. Actually, the received signals often have low SNR, and the seized signals have short duration, which results in insufficient samples. Thus, the recognition rate is low when using the existing methods. Therefore, this paper mainly researches the extraction technique of individual subtle features(signalprints) using the steady signal and aims at the problem of identifying individual transmitter with the same model.The basic theory of signalprints is researched. Based on signalprints mechanism, the characteristics of frequency, modulation parameters and stray output resulting from indi-vidual transmitter are researched, and methods for extracting signalprints are explored from different angles, such as time domain, frequency domain and high order spectra, etc. After that, an architecture of individual transmitter identification is established. The pro-posed algorithms are verified to be efficient using the measured radio data.The main areas in this paper are as follows.(1)The signalprints extraction based on carrier frequency and symbol rate is studied. An improved method of phase fitting is proposed to estimate the carrier frequency, and a method based on STFT time-frequency energy distribution and wavelet analysis is given for estimating the symbol rate of modulation signal in non-cooperative communication environment. Experimental results show that estimated frequency and symbol rate can be used as part of signalprints, and cooperate with other features for identifying individual transmitter.(2)The stray output of individual signal is used to extract signalprints. First, a method of orthogonal component reconstruction is presented for extracting signal envelop, and the fractal dimensions and Lempel-Ziv complexity are used to extract the spurious modulation features of signal envelop. Then the individual differences of Hilbert edge spectrum sym-metry parameter and Hilbert Huang Transform time-frequency distribution(HHT-TFD) grayed image are studied for classifying. Experiments show that complexity features of envelop and HHT-TFD grayed image have better separability in lower SNR, whereas spectrum symmetry features perform worse as affected by the estimation accuracy of car-rier frequency.(3)The high order spectrum is used to extract signalprints. The square integrated bis-pectra(SIB) are proposed, and an improved local linear embedded method is used to re-duce dimension of SIB features. Studies show that SIB features are superior to the other local bispectra when extracting signalprints, and the low-dimensional SIB features have good performance of clustering and noise suppresion.(4)Aimed at actual applications of transmitter identication, a multi-class SVM classi-fier based on kernel distance measurement and a combination classifier based on D-S evi-dence theory are proposed for verifying the effectiveness of signalprints extracted from different angles. Experiments show that the feature set of signalprints extracted from steady signal can achieve high recognition rate for measured radio data.This paper solves the problem of identifying individual transmitter with the same model under the conditions of small samples, lower and variant SNR to some extent. The research has some theoretical and practical application prospects.更多还原