【作者】 王宏；

【导师】 周正欧；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2010， 博士

【摘要】 超宽带穿墙雷达能够实现对建筑物或障碍物后面目标的探测、定位、成像和跟踪,在军事装备、城市安全、火灾及地震等自然灾害搜救方面有着广泛的应用前景。要实现穿墙雷达探测的良好性能,除了系统设计时考虑好雷达的各种参数,设计出性能良好的雷达系统外,信号处理也是一个非常重要的方面。因此,本论文对穿墙雷达信号处理的四个重要方面:目标定位、成像、多普勒频率特性分析和运动模式识别,进行了深入而细致的研究。本论文的主要创新之处在于:1、对于有墙体情况,提出了采用牛顿法求解非线性定位方程组和二维空间搜索法两种目标定位方法,给出了单发双收模式下穿墙雷达目标定位的仿真结果。分析了噪声对定位精度的影响,研究了两个墙参数(墙的介电常数和厚度)都未知情况下的墙参数估计和目标定位问题,给出了目标定位的仿真结果。2、针对有桌椅、柜子等家具的复杂背景房间中,采用后向投影(Back Projection,简称BP)成像,人等运动目标无法与背景区分开来的问题,提出一种动目标BP成像算法。对于超宽带脉冲穿墙雷达,将连续两帧对应接收天线的回波信号相减再BP成像,能够有效消除收发天线间的直耦、墙及墙后家具等静止背景杂波图像,保留人等动目标的像;对于超宽带噪声穿墙雷达,将连续两帧对应接收天线回波与发射噪声信号的互相关信号相减,也能够有效消除收发天线间的直耦、墙及墙后家具等静止背景杂波,保留人等动目标的回波信息。将相减后的信号(对脉冲雷达是回波信号之差,对噪声雷达是互相关信号之差)利用BP算法进行时域相干叠加,从而获得动目标人的像,有效提高了图像信杂比。对上述两种穿墙雷达体制,将生成的动目标图像相叠加,均可得到动目标的运动轨迹图。利用时域有限差分方法(Finite Difference of Time Domain,简称FDTD)模拟了多种穿墙成像场景,仿真结果表明,该方法行之有效,能够对人等动目标进行检测、定位、成像和跟踪。3、由于经验模式分解(Empirical Mode Decomposition,简称EMD)可能存在模式混合问题,并且整体平均经验模式分解(Ensemble Empirical Mode Decomposition,简称EEMD)对信号中的噪声抑制作用不明显,提出了一种改进的EEMD方法。将该方法应用到穿墙雷达人的运动多普勒特性分析中,并通过希尔伯特-黄变换(Hilbert-Huang Transform,简称HHT)得到多普勒信号的时间-频率-能量谱。仿真实例和实验结果分析均表明,改进的EEMD方法不仅能够有效地消除模式混合问题,将被测信号中体现不同运动细节的多普勒频移成分独立分解在不同的本征模式函数(Intrinsic Mode Function,简称IMF)中,而且还能够有效抑制原信号中的噪声,得到更清晰的时频分布。4、利用经验模式分解和支持向量机(Support Vector Machine,简称SVM)相结合的方法对穿墙雷达墙后人的多种运动进行模式识别。分别采用EMD、EEMD和改进的EEMD将穿墙雷达多普勒探测中人的静止站立、静止手臂摆动、前进后退一步、行走、跑步共5种运动的多普勒信号进行分解,选取各IMF能量占总IMF能量的百分比作为特征向量,对实验采集的564组数据采用支持向量机学习算法进行训练和识别,最大正确识别率达到94%以上。更多还原

【Abstract】 Because the ultra wide band (UWB) through wall radar has the ability of moving targets detection, localization, imaging and tracking behind the buildings or the barriers, it is widely used in military devices, city security, person secure at fire or earthquakes, etc. In order to obtain good performance, all kinds of key parameters should be considered seriously in the UWB through wall radar system design. Besides that, signal processing is also an important aspect. In this thesis, four important contents including targets localization, imaging, Doppler characteristic analysis and targets movement classification are studied. The main innovations of this thesis are as follows:1. For wall existing situation, two localization methods are proposed, which are the method of getting the numerical value of nonlinear localization equations using Newton method and the method of searching the position of the target in two dimentional geometry space. Simulation results of the target localization are given using the radar with one transmit antenna and two received antennas. Noise effect to the target localization precision is analyzed. The issues of wall parameters estimation and the target localization with unknown wall parameters (the wall thickness and dielectric constant) are discussed and simulation results are given.2. For a room with complex background such as full of furniture with desks and chairs, using the back projection (BP) algorithm the moving target of the human can not be distinguished from other background objects. To solve the problem, the moving target BP algorithm is proposed. For the UWB pulse through wall radar, using the BP algorithm to the difference signal obtained by subtracting the previous received signal from the current received signal, the image of background is eliminated greatly and the image of the moving target can be seen clearly. For the UWB noise through wall radar, by subtracting the previous frame cross-correlation functions between the received signal with the transmit noise signal from current frame cross-correlation functions, the coupling signal between the transmit and receive antennas as well as the background clutter including the wall and the furniture behind the wall are eliminated greatly and the reflection information of the moving target is reserved. Using the BP algorithm to the difference signal the cross-correlation functions of two successive frame samples, the image of the moving target is obtained and the signal to clutter ratio of the image is improved greatly. For above two radar systems, by refreshing difference images on the screen sequentially and adding all the images of the moving target together, we can get the track of the moving target. Many through wall senarioes are simulated using the finite difference time domain (FDTD) method for two radar systems. The simulation results show that the moving target BP algorithm is valid for human detection, localization, imaging and tracking in complex environment.3. Because the mode mixing problem may exist in the empirical mode decomposition (EMD) and the ensemble empirical mode decomposition (EEMD) method doesn’t have good effect on noise suppression in the signal, an improved EEMD method is proposed. This method is applied to the Doppler characteristic analysis of the human movements and the time-frequency-energy spectrum is obtained by Hilbert-Huang transform (HHT). Simulation and experimental results show that the improved EEMD method can not only eliminate the mode mixing problem by decomposing different Doppler frequency components to different intrinsic mode functions (IMFs), but also suppress the noise in the signal effectively so that more detail information can be seen clearly in the time-frequency spectrum.4. The classification of different kinds of movements of the human is realized by EMD combined with the method of support vector machine (SVM). Five types of movements of the human including standing still, standing with arms waving, one stepping forward and back, walking and running are analized respectively by EMD, EEMD and improved EEMD. The IMFs of each type of movement are obtained by three decomposition methods. Selecting the energy ratio of each IMF to the total IMFs as the feacture vector, using the SVM to the 564 group experimental samples, different types of movements of the human can be recognized and the biggest correct recognition rate is above 94%.更多还原