【作者】 柴娟芳；

【导师】 司锡才；

【作者基本信息】 哈尔滨工程大学 ， 通信与信息系统， 2009， 博士

【摘要】 为了提高被动雷达导引头在密集、复杂、多变的信号环境中正确分选跟踪信号的能力,本课题围绕雷达信号的分选识别算法在被动雷达导引头中的应用展开研究。在雷达信号分选跟踪器的软硬件设计与实现基础上,分析了传统分选模型存在的不足与缺陷,提出了一种改进的雷达信号分选模型。论文针对该改进模型的同时到达信号的分离、预分选参数的选择提取、脉内调制特征的选择提取等功能模块展开深入细致探讨,并对每个问题提出了相应的解决方法。同时,计算机仿真实验验证了各种方法的有效性和可行性。针对传统分选模型不能分选同时到达信号的缺陷,借助盲源分离技术研究了几种对同时到达信号分离具有明显优势的新方法。首先,将Fast ICA算法应用于分离雷达信号中,提出了基于Fast ICA算法的雷达信号分选算法,它可以很好地分离各种不同调制类型的脉冲雷达信号。其次,结合全局最优盲源分离算法,提出了基于伪信噪比最大化的盲源分离算法,该算法从独立信号完全分离时信噪比最大出发,建立基于源信号与噪声协方差矩阵的伪信噪比目标函数,并将目标函数的寻优过程转换为求解广义特征值的问题·,它不需要任何迭代运算,具有较低的计算复杂度,且信源独立就可以保证算法有解。然后,针对盲源分离开关算法无法有效分离多源信号的缺陷提出了盲源分离拟开关算法,它用峭度作判断函数自适应选择加权激活函数,该算法可以更加有效地分离空间未知多源线性混叠信号。针对单一相位差提取需要较高信噪比的缺陷以及天线往多元化和立体化方向发展的现象,提出了基于纯相位向量的动态聚类预分选算法。它先利用宽带数字信道化接收机提取多个通道间的相位差构造纯相位向量或纯方向角向量,并以此作为聚类对象,采用准C-均值动态聚类法和序列搜索法完成对空间雷达信号的预分选处理。仿真实验表明该算法能更好地实现较低信噪比下的预分选处理。在分析极化特征可作为信号分选参数的基础上,针对现有宽频带被动雷达导引头无法提取极化特征的缺陷,提出了基于双极性曲折臂天线的极化干涉仪的分选支路设计,并给出了基于数字信道化技术的极化参数频域提取的具体模型。将极化分选的概念首次引入被动雷达导引头的信号分选中,为分选同时到达、同频、同向等复杂雷达信号提供全新可靠的途径。针对经典参数无法正确描述脉内调制雷达信号的问题,从硬件可实现和理论前沿研究角度分别提出了基于IF子代特征向量的提取与自动分类算法和基于FRFTα域-包络曲线的特征向量提取与聚类分选算法。前一种方法首先对瞬时频率作相关预处理,提取瞬时频率方差、相关系数及其自相关函数极值个数等多个参数构造子代特征向量,再构造自动分类决策树完成对空间雷达信号的分选；后一种方法先通过FRFT后搜索得到旋转角域的包络函数,提取包络曲线峰值所对应的a值、峰值大小以及包络曲线的峰度等三个参数构造新的特征向量,再结合聚类分选就所提取的新特征向量完成对各种雷达信号的分选。计算机仿真结果证实了上述两种新特征向量作为信号分选参数补充的有效性和可行性。最后,结合数字信道化接收机与分选处理器实际调试所遇到的问题,给出了基于数字信道化的新型分选具体模型及新分选方法。更多还原

【Abstract】 Radar signal sorting and recognition algorithm applied to passive radar seeker (PRS) is investigated in this dissertation, which is in order to improve the signal sorting and tracking capability of PRS in the dense, complicated and variational environments.The limitations of traditional signal sorting model are analyzed according to the design and implementation of radar signal sorting and tracking device in both software and hardware. And then an improved radar signal sorting model is developed. The functional modules of the improved model are discussed in detail, including the separation of simultaneous arriving signals, the extraction of pre-sorting parameters and the extraction of intra-pulse modulated feature of radar signals. Correspondingly, some solutions are provided. Meanwhile, simulations prove the efficiency and feasibility of these methods.Some new ideas and some novel methods based on blind signal process (BSP) are shown to dispose the simultaneous arriving signals. Firstly, a radar signal sorting algorithm based on the Fast Independent component analysis (Fast ICA) is given in the dissertation. This sorting algorithm can separate different modulated radar signals efficiently. Secondly, combined with global optimal blind source separation (BSS) algorithm, a new BSS algorithm is presented based on maximum pseudo-Signal Noise Ratio (SNR). The pseudo-SNR function, built as the objective function, is constructed by the covariance matrix of source signals and noise. This idea is formed based on the theory that SNR is maximal when source signals of statistical independence are completely separated. Then unmixing matrix could be obtained without any iteration, when the optimization of the objective function is transformed into solving Generalized Eigenvalue (GE) problem. This new algorithm is global optimal with low computational complexity, and the statistically independence of source signals can guarantee a feasible solution. Thirdly, a quasi switching algorithm of blind source separation is proposed based on switching algorithm. It uses the signal kurtosis as the judgement function, which is utilized to choose and weight the corresponding activation function adaptively. Compared with the original algorithm, this novel algorithm can be more effective in unknown multi-source separation of linear mixed signals.A new dynamic clustering pre-sorting algorithm is illuminated based on phase-only vector, considering that a single phase is trustless under the influence of noise and microwave antennas are multi-baseline or tri-dimensional-baseline, Firstly, it uses multiple channels of phases measured by broadband digital channelization receivers to construct the phase-only vector or AOA-only vector, which acts as clustering object. Afterwards, quasi-C-means dynamic clustering and sequence searching are adopted to complete the pre-sorting work to various radar signals. Simulation experiments show that, the dynamic clustering pre-sorting algorithm based on phase-only vector can achieve better performance than single phase method in low SNR.A spur track design of signal sorting is conceived based on polarization interferometer constructed by dual-polarized sinuous antenna (DPSA), due to polarization characteristic can be regarded as the attribute of radar signals, And an extraction of polarization parameters in frequency domain is constructed based on the digital channelization technology. The concept of polarization sorting is introduced into signal sorting of passive radar seeker firstly. Polarization sorting provides a brand-new and reliable approach to signal sorting in complicated environment, especially in sorting the signals with the same time arriving, signals with the same frequency and signals with the same phase.There are two feature extraction methods of intra-pulse modulated radar signal presenting in this dissertation, due to classical parameters can’t describe the intra-pulse modulated radar signal accurately. One is the feature extraction method based on the filial generation of Instantaneous Frequency (IF); the other is the feature extraction method to envelope function of rotation angle a domain of radar signals based on Fractional Fourier Transform (FRFT). The first method pretreats the IF sequence of different modulated signals, and then extracts standard deviations, correlation coefficients with sample sequence and the extremums number of auto-correlation function to construct filial generation vector. Afterwards, it constructs an auto-sorting decision-making tree to verify the sorting performance of filial generation vector. The second method searches the envelope function of rotation angle domain based on FRFT firstly, extracts the rotation angle value a of envelope function’s peak, the peak value and the kurtosis of the envelope function to construct a new feature vector. At the same time, a cluster sorting method based on the new feature vector is used to complete the radar signals sorting work. Computer simulation results verify the feasibility and effectiveness of these new feature vectors as the complement to classical parameters.Finally, combined with the problems in debugging digital channelization receiver and sorting processor, a new sorting model and method based on digital channelization technology is provided.更多还原