【作者】 刘涛；

【导师】 龚耀寰；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2010， 博士

【摘要】 超视距雷达的信号处理一直以来都是研究的热点,也面临很多技术难点。在超视距雷达的干扰中,尤以瞬态干扰对超视距雷达的影响较大。瞬态干扰主要包括无线电射频干扰、大气中的闪电干扰、流星余迹干扰回波等等。依靠自适应波束形成技术虽然能抑制部分干扰,但仍有残留的干扰影响,造成多普勒域中噪声基底的抬高,减低了目标的检测性能,甚至可能造成目标完全淹没在噪声中。本文在研究其他学者研究成果的基础上,深入研究了波束形成后的时域信号抗瞬态干扰算法。本论文的主要工作和创新点主要体现在以下几个方面:1.继承并改进已有算法在参考以前学者的研究成果基础上,遵循经典的“杂波抑制-时域干扰检测-挖除-恢复”进行了抗干扰框架的建立,其中实现了小波检测算法。将已有文献中小波系数的计算方法改为基于CZT的快速算法,并详细讨论了其中的实现细节。2.提出两种新的干扰检测算法第一种算法基于分形方法,实现了在杂波信号中直接检测瞬态干扰的方法。第二种算法借助回波信号的短时信号子空间结构具有时变特性来检测干扰,且处理前不需要抑制杂波。3.提出广义频域概念和模型将瞬态干扰建模为一个理想Dirac函数,其多普勒谱建模为一个广义正弦信号。这个广义正弦信号的频率正好对应了干扰的时域位置。由此,引出了一大类瞬态干扰的检测算法。本文使用MUSIC算法和AR模型算法,成功检测了瞬态干扰的时域位置,且只需要采用最简单的理想高通滤波器抑制杂波。基于广义正弦信号模型和子空间投影算法,提出将干扰检测和干扰抑制进行联合处理:在理想高通滤波滤除杂波后,采用主成份分析针对广义正弦信号进行分析,确定瞬态干扰的广义正弦信号对应的特征值和特征矢量,并重构其子空间(将其定义为“信号子空间”)。将含有杂波的多普勒谱信号投影到“信号子空间”的正交补空间上(即“噪声子空间”),就完成了滤除瞬态干扰的目的。4.建立软硬件仿真平台为了克服实测数据较少的难题,建立了抗瞬态干扰仿真平台提供大量数据验证算法。建立了基于MATLAB的软件仿真平台,并引入GPU通用计算,使用显卡作为仿真平台的硬件,大幅度提升了MATLAB仿真效率,而无需转换到DSP就能获得很高的计算效率。更多还原

【Abstract】 The signal processing of Over-the-horizon radar (OTHR) is always one of the research hot-spots, and faces many technical difficulties. Among those interferences in OTHR, the impulsive interference (IMI) has the great impact. The IMI mainly include the radio frequency interference, the lightning in the air, and the meteor echo in the universe. Although the adaptive digital beamforming (ADBF) can suppress these IMIs to a certain extent, but there is still some remained interference impact, lifting up the noise bases and depressing the detection performance of target peak, even submerging the target into the noise background .Based on the fruits of other previous researchers, this dissertation makes some research on algorithms for OTHR temporal signal anti-impulsive-interference after digital beamforming. The main works and the innovation points include:1. Inherited and improved the existing algorithmsBased on the existing fruits of other researchers, it firstly follows the classical“clutter suppression”-“temporal detection”-“restoration”routine, and construct the framework for the anti-impulsive-interference tasks. It realizes wavelet detection for IMI detection. It changes the existing algorithms for the wavelet coefficient in the previous papers into the fast algorithm based on CZT, and discusses its details.2. Proposed two new detection algorithmThe first one is based on the fractal method, and realizes the interference detection in the clutter background. The second one relies on the time-varying characteristics of short-time signal subspace to detection the position of interference, without clutter suppression before detection.3. Propose the concept and model of the general frequency domainThe impulsive interference is modeled as an ideal Dirac function, whose Doppler spectrum is modeled as a general sinusoid signal. The frequency of this general sinusoid signal is just corresponding to the IMI temporal position. Thus, it leads to a batch of detection algorithms. This dissertation utilizes MUSIC algorithm and AR model algorithm to detect IMI sucessifully, only depending on the simplest ideal high-pass filter to suppress the clutter.Based on the above general sinusoid signal model and applying the subspace projection processing, this dissertation prososes a joint detection and suppression algorithm: after the ideal high-pass filtering, the principal component analysis (PCA) is processed on the genral sinusoid signal to determine the eigenvalues and eigen-vectors corresponding to the general sinusoid signal. These eigen-vectors can reconstruct the subspace of general sinusoid signal (referred to as signal subspace). The raw Doppler spectrum signal without clutter-suppression is projected into the orthogonal complement subspace of signal subspace (referred to as noise subspace), which can reaches the goal of filtering impulsive interference.4. Construct the software and hardware simulation platformTo solve the lack problem of echo signal in the research, it constructs the simulation platform to supply plenty of exeriamental data for algorithm verification. It constructs the software simulation platform based on MATLAB and hardware simulation platform based on GPU general computation, which greatly improves the efficiency of MATLAB compuatation instead of code transplantation into DSP.更多还原