【作者】 张高毅；

【导师】 唐斌；

【作者基本信息】 电子科技大学 ， 信息获取与探测技术， 2009， 博士

【摘要】 电波传播过程中会出现反射和散射等现象，使得空间信号环境变得更为复杂，传统的点源模型无法很好地描述这种信道模型，需要用分布源来描述。根据反射和散射面的粗糙等情况，分布源可分为相干分布源和非相干分布源。迄今为止，已提出了多种空间分布源模型和相应的波达方向估计方法，但其中多数方法需要多维搜索，算法复杂度高。此外，尚未见考虑多种分布形式并存时的信号模型及相应的参数估计方法。本论文对两种分布源进行了深入研究，提出了相应的参数估计方法。在此基础上，研究了多种形式信号源并存时的信号模型及相应的参数估计方法。主要研究成果如下：1、研究了相干分布源的高斯近似模型。根据相干分布源的近似模型可知，在小角度扩展情况下，角度扩展主要影响广义方向向量各元素的幅度。可以从广义方向向量中提取出相干分布源的相位信息以估计波达方向，从广义方向向量的幅度信息中估计角度扩展。据此分别提出了基于高阶累积量、矩阵束和二次虚拟内插的相干分布源二维波达方向估计方法。根据高阶累积量可扩展阵元的特性，利用较少阵元即可实现多个信号的参数估计。基于矩阵束的方法则利用其可提取出广义方向向量的相位信息的原理，可有效的估计相干分布源的波达方向。基于二次虚拟内插的方法可有效提高内插精度，提高相干分布源的参数估计精度。2、研究了非相干分布源的空间频率近似模型，根据非相干分布源的近似模型可知，在小角度扩展情况下，角度扩展主要影响阵列协方差矩阵各元素的幅度。可以从非相干分布源的协方差矩阵各元素的相位信息中估计波达方向，从协方差各元素的幅度信息中估计角度扩展。据此提出了基于协方差矩阵重构的非相干分布源波达方向和角度扩展估计方法。该方法可有效地实现非相干分布源波达方向和角度扩展的分离估计。在估计波达方向时，分别提出了利用最小均方拟合方法和MUSIC类方法两种算法。通过空间三维阵列结构设计，将MUSIC类方法成功推广到非相干分布源的二维波达方向估计中。3、研究了点源、相干分布源和非相干分布源共存的复合源信号模型，根据角度扩展对相干分布源广义方向向量以及非相干分布源协方差矩阵元素的影响，提出了基于SOBI的复合源信号类型盲辨识及相应的波达方向估计方法。该方法可有效地对信号源进行辨识并实施分离，在此基础上，通过利用MUSIC类方法和匹配跟踪方法，可实现各信号源的波达方向有效估计。4、研究了存在散射情况下的信号多维参数估计问题。根据分布源信号模型，提出了基于PRO-ESPRIT算法的多普勒频率、方位角和俯仰角的多维参数估计方法。该方法通过对阵列信号的延时，将信号的多普勒频率信息提取出来，这样可从广义方向向量的相位信息中获得相干分布源的方位角和俯仰角估计。更多还原

【Abstract】 Wave propagation may cause reflection and scattering phenomenon, makes thespacial signal environment become complicated, the channel model can’t be describedappropriately using the traditional point source model, whereas distributed source modelis needed. According to the roughness of the reflection or scattering surface, distributedsource can be classified into coherently distributed (CD) source and incoherentlydistributed (ICD) source. Up till now, numerous spatial distributed source models andcorresponding direction of arrival (DOA) estimation algorithms were proposed.However, most of the existing algorithms need multiple dimensional searches, thecomputation burden is heavy. What’s more, a model which describes the coexistence ofmultiple types of sources and corresponding parameter estimation algorithm is not seenyet. This dissertation studied the two types of distributed source deeply, thecorresponding parameter estimation algorithm is proposed. Based on this, the signalmodel which describes the coexistence of multiple types of sources and correspondingparameter estimation algorithm is proposed. The main contributions of this dissertationare as follow.1. The Gaussian approximation model for coherently distributed source is studied.According to the approximation model for CD source, we know that, in thecircumstance of small angular spread, the angular spread mainly affects theamplitude of the general steering vector (GSV). We can extract the phaseinformation of the GSV to estimate the DOA of CD source, and estimate the angularspread of CD source from the amplitude information of the GSV. The correspondingalgorithms which based on cumulants, matrix pencil and twice virtual interpolationsmethod were proposed. According to the character of the cumulants that the arraysensors can be spread, we can realize the parameter estimation of multiple signalswith quite little sensors. The method based on matrix pencil used the character thatthe phase information of the GSV can be extracted from the received data, thusDOA for CD source can be estimated effectively. The method based on twice virtualinterpolations can improve the precision of the interpolation, thus improve the estimation precision of the parameters of CD sources.2. The spatial frequency model for incoherently distributed source is studied.According to the model, when the angular spread is small, the angular spread mainlyaffects the amplitude of the elements of the covariance matrix. We can estimate theDOA of incoherently distributed source from the phase of the elements of thecovariance matrix, and estimate the angular spread from the amplitude of theelements of the covariance matrix. Thus the DOA and angular spread estimationmethod for incoherently distributed source is proposed based on the reconstruct ofthe covariance matrix. This method can realize the separable estimation of DOA andangular spread for incoherently distributed source. When estimating the DOA, thealgorithms were proposed based on least square fitting and MUSIC-like algorithm,respectively. By designation of the spatial 3D array structure, the MUSIC-likealgorithm is extended to the 2D-DOA for incoherently distributed source.3. The composite signal model which describes the coexistence of point source,coherently distributed source and incoherently distributed source is proposed.According to the affection of the angular spread to the coherently distributed sourceand incoherently distributed source, the blind identification and DOA estimationmethod for composite source is proposed. The method can identify the sources andseparate each source effectively, based on this, by using MUSIC-like method ormatching pursuit method, the corresponding DOA for each source can be estimatedeffectively.4. The estimation of multiple dimensional parameters for signal sources in presence ofscattering is studied. According to the distributed source signal model, the multipledimensional parameters estimation method for the estimation of Doppler frequency,azimuth and elevation is proposed based on PRO-ESPRIT. This method can extractthe Doppler frequency information from the signal by using time delay of the arraysignal, thus we can estimate the azimuth and elevation of coherently distributedsource from the phase information of the GSV.更多还原