混合高斯概率密度模型动态簇算法参数估计

混合高斯概率密度模型可以很好地拟合样本的概率密度。在各高斯分量概率密度互不重叠的条件下，使用动态簇算法可以快速而精确地估计出混合高斯概率密度模型参数。这是一种基于最小均方差原则的递推算法，在正向推导出各种可能簇边界后，再根据确定的最末边界值逆向推定各前导簇边界，从而得到混合高斯概率密度模型参数估计值。算法介绍之后，给出了两个拥有不同概率密度分布的仿真建模实例．最后总结分析了该算法的优劣，并简介了算法的推广．     

混响的混合高斯概率密度建模

混合高斯模型能够有效地拟合混响背景的一维概率密度分布。常用的混合高斯概率密度模型参数估计方法是EM迭代算法,但这种算法的主要缺点是估计精度过分依赖于初始值。而GreedyEM算法通过往混合模型中不断地加入高斯分量,能很好地解决这一问题。文章将多维图象处理中的GreedyEM算法加以合理简化,并给出模型自动定阶方法,从而成功应用于水声混响的一维混合高斯模型建模中。实验结果表明:应用新算法能从混响接收数据中准确拟合其概率密度曲线,并且能适应不同的数据长度,具有很好的通用性。     

混合高斯概率密度模型参数的期望最大化估计

混合高斯模型是对非高斯数据进行概率密度拟合典型模型,其参数估计可以通过期望最大化(EM)迭代算法获得。多维混合高斯模型参数的EM估计因结构庞杂而难以求解,而对主动检测背景的统计特性拟合来说,一维的混合高斯模型一般即已足够。描述了该情形下的混合高斯模型及其参数估计问题之后,导出了一种工程实用的、简化的EM迭代算法,并给出了可计算机编程实现的算法流程图。然后详细探讨了对EM估计精度与速度有着重要影响的参数初始化问题,给出了三种可选择的初值设置方案:高速度方案、高精度方案和二者的折衷方案,并分析了它们各自的适用场合。最后,结合一组数值仿真实例,演示了EM迭代算法的良好的混合高斯模型参数估计性能。     

混合高斯参数估计的两种EM算法比较

混合高斯模型是一种典型的非高斯概率密度模型,获得广泛应用。其参数的优效估计可以通过最大似然方法获得,但最大似然估计往往因其非线性而难以实现,故期望最大化(Expectation-Maximization,EM)迭代算法成为一种常用的替代方法。常规EM算法性能受迭代初值设置影响大,且不能对模型阶数做出估计。一种名为贪婪EM的改进算法可以克服这两个缺点,获得更为准确的模型参数估计,但其运算量一般会远大于前者。本文对这两种EM算法进行综合研究,深入挖掘两者之间的关系,并基于相同的数值仿真实例,直观地演示比较两者的性能差异。     

混响背景下GTM回波检测算法研究

为改善混响背景下传统匹配滤波算法效果不佳问题,在分析其非平稳性、有色性和非高斯性的基础上,提出了混合高斯时变自回归模型(Gaussian mixture Tvar Model,GTM),推导了模型公式及其参数求解方法,形成了GTM回波检测算法。为对混响特性及滤波效果进行定量描述进而验证算法性能,给出了一种定量衡量混响非平稳性、有色性、非高斯特性的滤波效果评价方法。通过实测混响分析表明,GTM模型能够较好地拟合实测混响的概率密度曲线和功率谱密度曲线,实现了混响背景下回波的有效检测并改善混响特性。     

基于L型线阵的二维分布源参数估计方法

针对二维分布源信号波达方向估计问题,提出了一个新的二维分布源模型及其参数估计方法。将基于角度信号密度函数的一维相干分布源模型扩展至二维,推导了二维分布源的广义阵列方向向量。在L型线阵下,利用广义MUSIC算法,首先估计出俯仰角及其扩散参数,然后利用估计出来的俯仰角及扩散参数对方位角及扩散参数进行估计。仿真实验表明,所提出的二维相干分布源参数估计算法具有良好的定向精度。     

偏斜非高斯随机振动信号幅值概率密度函数研究

偏斜非高斯振动信号幅值概率密度没有明确、简洁的解析表达式。研究概率密度的解析表达式,对于非高斯振动理论研究具有重要意义。针对以上需求,提出了一种基于高斯混合模型的概率密度函数表示方法。首先,通过时间样本序列得到偏斜非高斯振动信号前五阶矩的估计值。其次,根据平稳高斯随机过程各阶矩之间的定量关系,结合二阶高斯混合模型的数学表达式建立方程组,求解得到混合模型中每个高斯分量的均值、标准差和权重系数。然后,将每个高斯分量的参数代入高斯混合模型,得到偏斜非高斯振动信号的幅值概率密度的解析表达式。最后,将所提出的方法应用于仿真非高斯加速度信号和实测非高斯振动应力信号,充分验证了该方法的有效性和适用性。     

基于多模型自适应技术的信道估计研究

提出了一种基于多模型自适应技术的信道估计方法.该方法通过设计动态信道模型库来覆盖未知信道的状态的改变,基于模型切换和动态规划(DP)进行模型变化检测及信道参数估计,从而在变化的信道环境中将有限个简单子集的估计算法有机地结合成一种在大信道范围内具有高度鲁棒性的信道估计器.仿真实验结果表明,此信道估计器比自适应信道估计器方法具有优势.     

混合广义线性模型的统计推断

为了更好地拟合实际数据,本文提出了混合广义线性模型并进行参数估计．首先,基于异质总体的一阶矩以及二阶矩存在的条件下,运用混合广义线性模型对子总体的均值进行建模,构造扩展拟似然和伪似然函数,然后利用 EM 算法对均值参数、散度以及混合比例进行估计,并通过 Monte Carlo 模拟验证所提出的模型参数估计方法的有效性．最后,实例研究的结果表明本文的模型和方法是科学实用的．     

考虑风向风速联合分布的输电塔线体系风振疲劳研究

输电塔线体系属于风敏感结构体系,在结构设计和实际工作中风荷载均起控制作用。然而,在风荷载计算以及工程研究领域中对风向风速的处理至关重要,但因两者之间的相关性导致了建立联合分布模型的困难。为此,该文基于乘法定理以重庆市日极值风速为对象,建立了风速条件概率密度的混合模型,改善了单一概率分布模型的不足。与此同时,根据风向方位记录数据建立了风向角的概率密度函数模型,并结合风速条件概率密度的混合模型,给出了风向风速的离散-连续混合联合分布模型。然后,将风向风速联合分布模型与Miner线性疲劳累积损伤理论相结合,推导出了输电塔的风振疲劳计算方法。最后,采用此建议算法对特高压输电塔线结构体系进行了风振疲劳分析,有效地考虑了风向风速对输电塔体风振疲劳的影响,减少了计算量,提高了结构分析的计算效率。     

Azimuth-invariant bistatic multichannel synthetic aperture radar for moving target detection and location

Ground moving target indication (GMTI) is one of the most important applications of the bistatic synthetic aperture radar (SAR) system as well as the monostatic system. An algorithm for moving target detection and location is presented with an azimuth-invariant bistatic multichannel SAR, which consists of one transmitter (channel) and multireceivers (multichannel). The algorithm is based on the discussion of the particularities of the bistatic SAR configuration including coherence improvement and clutter characteristics. Then, the corresponding compensating methods including two-dimensional range-azimuth prefiltering and bistatic differential range correction are proposed to solve these particularities. It is shown that using the compensating methods, the stationary clutter can be suppressed and the moving parameters of ground targets can be estimated accurately. Finally, simulation results demonstrate the effectiveness of the proposed algorithm.     

Optimal multichannel estimation of the location of a target with nonoverlapping noise

We propose a new algorithm for estimating the location of an object in multichannel images when the noise is spatially disjointed from (nonoverlapping with) the target. This algorithm is optimal for nonoverlapping noise and for multichannel images in the maximum-likelihood sense. We consider the case in which the statistical parameters of the input scene are unknown and are estimated by observation. We assess the results for simulated images with white and Gaussian background, for a large scale of variances of the background noise, and different values of the contrast in the scene. We compare the results of this algorithm with the results obtained with two other algorithms, the optimal algorithm for monochannel nonoverlapping noise and the optimal algorithm for multichannel additive noise, and we show that in both cases improvement can be obtained. We show the efficiency of the estimation for real input scenes when the background noise is correlated clutter noise. This algorithm has the same complexity as correlation, and the improvement is obtained with no more calculation cost than with classic methods.     

Three-dimensional channel modelling using spherical statistics for multiple-input multiple-output systems

Recently, the third generation partnership standards bodies (3GPP/3GPP2) have defined a two-dimensional channel model for multiple-input multiple-output (MIMO) systems, where the propagating plane waves are assumed to arrive only from the azimuthal direction and therefore not include the elevation domain. As a result of this assumption, the derived angle-of-arrival (AoA) distribution is characterised only by the azimuth direction of these waves. The AoA distribution of multipaths is implemented with a novel three-dimensional approach. The von Mises- Fisher (VMF) probability density function is used to describe their distribution within the propagation environment in both azimuth and co-latitude. More specifically, the proposed model uses a mixture of VMF distributions. A mixture can be composed of any number of clusters and this is clutter specific. The parameters of the individual cluster of scatterers within the mixture are derived and an estimation of those parameters is achieved using the spherical K-means algorithm and also the expectation maximisation algorithm. Statistical tests are provided to measure the goodness of fit of the proposed model. The results indicate that the proposed model fits well with MIMO experimental data obtained from a measurement campaign in Germany.     

Statistically tuned Gaussian background subtraction technique for UAV videos

Background subtraction is one of the efficient techniques to segment the targets from non-informative background of a video. The traditional background subtraction technique suits for videos with static background whereas the video obtained from unmanned aerial vehicle has dynamic background. Here, we propose an algorithm with tuning factor and Gaussian update for surveillance videos that suits effectively for aerial videos. The tuning factor is optimized by extracting the statistical features of the input frames. With the optimized tuning factor and Gaussian update an adaptive Gaussian-based background subtraction technique is proposed. The algorithm involves modelling, update and subtraction phases. This running Gaussian average based background subtraction technique uses updation at both model generation phase and subtraction phase. The resultant video extracts the moving objects from the dynamic background. Sample videos of various properties such as cluttered background, small objects, moving background and multiple objects are considered for evaluation. The technique is statistically compared with frame differencing technique, temporal median method and mixture of Gaussian model and performance evaluation is done to check the effectiveness of the proposed technique after optimization for both static and dynamic videos.     

Automated defect detection system using wavelet packet frame and Gaussian mixture model

This paper proposes an approach for automated defect detection in homogeneous textiles using texture analysis. The texture features are extracted by the wavelet packet frame decomposition followed by the Karhunen-Loève transform. The texture feature vector for each pixel is used as an input to a Gaussian mixture model that determines whether or not each pixel is defective. The parameters of the Gaussian mixture model are estimated with nondefective textile images in supervised defect detection. An approach for unsupervised defect detection is also presented that can identify the heterogeneous subblocks on the basis of the Kullback-Leibler divergence between two Gaussian mixtures. The proposed method was evaluated on 25 different homogeneous textile image pairs, one of each pair with a defect and the other with no defect, and was compared with existing methods using texture analysis. The experimental results yielded visually good segmentation and an excellent detection rate with a low false alarm rate for both supervised and unsupervised defect detection. This confirms the validity of the proposed approach for automated defect detection and localization.     

Comparison of Dynamic Differential Evolution and Self-Adaptive Dynamic Differential Evolution for Buried Metallic Cylinder

The application of two techniques for the reconstruction of shape reconstruction of a metallic cylinder from scattered field measurements is studied in this paper. These techniques are applied to two-dimensional configurations, for which the method of moment (MoM) is applied to solve the integral equations. Considering that the microwave imaging is recast as a nonlinear optimization problem, an objective function is defined by the norm of the difference between the measured scattered electric fields and those calculated for each estimated metallic cylinder. Thus, the shape of a metallic cylinder can be obtained by minimizing the objective function. In order to solve this inverse scattering problem, two techniques are employed. The first one is based on dynamic differential evolution (DDE) algorithm, while the second one is an improved version of the DDE algorithm with self-adaptive control parameters, called SADDE. Both techniques are tested for the simulated data contaminated by additive white Gaussian noise. Numerical results indicate that SADDE algorithm outperforms DDE algorithm in terms of reconstruction accuracy and convergence speed.     

Real-time adaptive clutter rejection filtering in color flow imaging using power method iterations

We propose a new algorithm for real-time, adaptive-clutter-rejection filtering in ultrasound color flow imaging (CFI) and related techniques. The algorithm is based on regression filtering using eigenvectors of the signal correlation matrix as a basis for representing clutter, a method that previously has been considered too computationally demanding for real-time processing in general CFI applications. The data acquisition and processing scheme introduced allows for a more localized sampling of the clutter statistics and, therefore, an improved clutter attenuation for lower filter orders. By using the iterative power method technique, the dominant eigenvalues and corresponding eigenvectors of the correlation matrix can be estimated efficiently, rendering real-time operation feasible on desktop computers. A new adaptive filter order algorithm is proposed that successfully estimates the proper dimension of the clutter basis, previously one of the major drawbacks of this clutter-rejection technique. The filter algorithm performance and computational demands has been compared to that of conventional clutter filters. Examples have been included which confirms that, by adapting the clutter-rejection filter to estimates of the clutter-signal statistics, improved attenuation of the clutter signal can be achieved in normal as well as more excessive cases of tissue movement and acceleration.     

晃动液体单摆模型动力学参数的频域辨识

本文研究晃动液体动力学参数的频域辨识法,详细推导了晃动液体的等效单摆模型,研制了贮箱模型、悬挂装置和高灵敏度测力计。通过测量贮箱位移和激振力,求得其位移对激振力的频率特性数据。首先用误差校正法辨识传递函数系数,然后用数学规划法确定模型的动力学参数。同时,通过测量晃动液体的扰动力矩,直接推算单摆模型的悬挂点位置。辨识出的单摆模型动力学参数值与按位势理论计算的数值接近。研究结果表明本文采用的频域辨识法有效。