中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
微弱信号检测与基于阵列的信源定位新方法研究

文献类型:学位论文

作者梁军利
学位类别博士
答辩日期2007-06-08
授予单位中国科学院声学研究所
授予地点声学研究所
关键词微弱信号检测 定位 跟踪
其他题名Study of new approaches to weak signal detection and source localization
学位专业信号与信息处理
中文摘要微弱信号检测、信源定位及目标跟踪是信号处理领域的重要研究内容,其在鱼雷、声纳、雷达等领域有着广泛的应用。如何抑制噪声(混响)提高信噪比(信混比)是微弱信号检测的关键步骤,减少阵列孔径损失、避免参数配对、提高参数估计精度是信源定位的难点所在,而借助状态空间方程高精度地递归估计目标状态是目标跟踪系统的重要环节之一。本论文的研究涉及这三个方面的内容。自适应陷波器是抑制窄带混响、检测微弱信号的关键技术。本论文利用公式推导得出自适应混响中心频率估计的状态空间方程表示;然后基于“自校正”的思想,得出一种无须观测噪声方差先验信息的自适应卡尔曼滤波算法。该算法收敛速度快,能够同时完成估计混响中心频率和滤除混响的功能,可以直接在数据域执行,计算复杂度低。并和其他陷波器进行了比较。然后用Moose混响模型产生仿真数据,进行了仿真数据的处理研究。 利用随机共振滤波器强有力滤除噪声的能力,结合正弦和线性调频信号在时频域内谱的特征信息,研究了被动方式下强噪声背景中微弱正弦和线性调频信号的检测。针对近场窄带信源的定位问题,提出了阵列孔径利用率高、参数估计精度高、无须参数配对的信源定位新算法。主要贡献是:针对近场窄带信源传播模型,利用高阶累积量的阵列扩展能力,计算一些高维矩阵,在高阶累量域形成平行因子分析模型,然后分析了该模型具有低秩分解的唯一性,最后从分解矩阵中联合估计信源参数。针对远场窄带信源的定位问题,分析了现有方法阵列孔径损失较大和需要参数配对的缺点,提出了一种阵列孔径利用率高、参数估计精度高、无须参数配对的新算法。主要贡献是:设计了有效的阵列结构,利用高阶累积量的阵列扩展能力,计算一些高维矩阵,在高阶累量域形成平行因子分析模型,然后分析了该模型具有低秩分解的唯一性,最后从分解矩阵中联合估计信源参数。针对目前一些借助状态空间方程递归估计目标状态的算法在非高斯非线性情况下性能严重下降的问题,提出了一种改进的粒子滤波算法。主要贡献是有效设计了重要采样函数,进而嵌入到经典的粒子滤波算法提高状态的估计精度。
英文摘要Weak signal detection, source localization, and target tracking are key problems in signal processing areas, such as torpedo, sonar, and radar systems. This thesis mainly deals with these problems, i.e. suppressing noise (reverberation) and improving the Signal-to-Noise-Ratio (SNR, or Signal-to-Reverberation-Ratio, SRR) in the signal detection problem, alleviating loss of the array aperture, avoiding pairing parameters, and improving the estimation accuracy in the source localization problem, and recursively estimating target state by virtue of state-space equations in target tracking problemAdaptive notch filter is one of the key techniques of suppressing reverberation and detecting weak signal. This thesis derives state-space equations for estimating the center frequency of reverberation adaptively, and proposes an adaptive Kalman filtering algorithm without the apriority of measurement noise, which can simultaneously estimate frequency and suppress reverberation, and has the characteristics of low computational complexity and direct application in data domain. In addition, the performance of this notch filter with others is compared with others, a simulation model of underwater acoustic channel and reverberation is introduced, and the simulation data based on this model is processed to show the performance of the proposed algorithm.Based on the signal feature in the time-frequency domain, a new algorithm, which efficiently uses the ability of stochastic resonance filter to suppressing heavy noise and, is proposed to detect the weak signals in heavy noise.A new algorithm without pairing parameters, which has the characteristics of both efficient utilization of array aperture and high estimation accuracy, is proposed to localize narrowband sources in the near field. This algorithm makes use of the ability of higher-order statistics to increase array aperture, forms high-dimensional cumulant matrices by virtue of some properly chosen cumulant, and constructs a parallel factor model in the cumulant domain. Finally, it estimates source parameters from the matrices, which are obtained from the low-rank decomposition of this model.Many available methods of localizing far-field narrowband sources suffers heavy loss of the array aperture and requires an extra procedure to pair parameters. To avoid these shortcomings, a new algorithm is proposed to jointly estimate frequencies and two-dimensional directions of arrival of far-field sources. This algorithm proposes a particular array configuration, and adopts the same techniques as those of near-field source case, such as cumulant and parallel factor model.Many algorithms by virtue of state-space equations have poor performance in the non-gaussian and nonlinear cases. To improve it, this thesis proposes a modified particle filtering algorithm, which improves the estimation accuracy by introducing a new sampling function.
语种中文
公开日期2011-05-07
页码129
源URL[http://159.226.59.140/handle/311008/168]  
专题声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文
推荐引用方式
GB/T 7714
梁军利. 微弱信号检测与基于阵列的信源定位新方法研究[D]. 声学研究所. 中国科学院声学研究所. 2007.

入库方式: OAI收割

来源:声学研究所

浏览0
下载0
收藏0
其他版本

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。