A NEW METHOD FOR EIGENSTRUCTURES EXTRACTION AND ITS NEURAL NETWORKS IMPLEMENTATION

The cost function for eigenstructures extraction is discussed in detail in this paper, one can obtain the largest eigenvector by minimizing the cost function. In order to obtain other eigenvectors, a covariance matrix series is constructed. If one compares the cost function with the energy function of a neural networks, the neural networks can be easily introduced to extract the eigenvectors. Theoretical analysis and computer simulations show that the proposed method is reasonable and feasible.     

A NEW METHOD BASED ON ESPRIT FOR ESTIMATING QUADRATIC PHASE COUPLING

This paper presents a new method based on ESPRIT for estimating the quadrati-cally coupled frequency pairs (QC pairs). This method constructs an ESPRIT matrix pencil by exploiting frequency symmetry property of complex third-order cumulants, QC pairs are yielded by evaluating the generalized eigenvalues of this matrix pencil. A simple analysis is provided to show the method is simpler. The accuracy and the high-resolution performance of this method are verified by simulations.     

A NEW ADAPTIVE FILTERING SCHEME BASED ON WAVELET TRANSFORM

Based on the scale function representation for a function in L2(R), a new wavelet transform based adaptive system identification scheme is proposed. It can reduce the amount of computation by exploiting the decimation properties and keep the advantage of quasi-orthogonal transform of the discrete wavelet, transform (DWT). The issue has been supported by computer simulations.     

DISTRIBUTED PARAMETER NEURAL NETWORKS FOR SOLVING PARTIAL DIFFERENTIAL EQUATIONS

Novel distributed parameter neural networks are proposed for solving partial differential equations, and their dynamic performances are studied in Hilbert space. The locally connected neural networks are obtained by separating distributed parameter neural networks. Two simulations are also given. Both theoretical and computed results illustrate that the distributed parameter neural networks are effective and efficient for solving partial differential equation problems.     

A NEW IMEMENTATION FOR DFT BASED ON DELTA MODULATION

This paper proposes a new implementation for DFT based on delta modulation which isdifferent from the conventional PCM implementation.The hardware structure of the method is multiplier-free,simple and low in cost.Theoretical analyses show that its SNR can approach conventional one's,forinstance,FFT.Computer simulations demonstrate that for deterministic signals,the results agree withtheoretical analyses.For handlimited Gaussian signals,we can still get results similar to the conventionalDFT's if selecting the step size suitably.A hardware structure which simply consists of ROM,an adder andother auxiliary circuits is also given.     

A NEW METHOD FOR SOLVING THE SCATTERING PROBLEM OF DIELECTRIC BODIES

A new approach to the EM scattering problem of an inhomogeneous lossy dielectric body is pro-posed.It is shown that the entire interior electric field distribution can be obtained from the measured exteri-or field distribution by simple recurrence relations.Detailed derivations of these recurrence relations for thefield distribution inside the scattering body are presented,and the results obtained by computer simulationsare given.     

A NEW METHOD FOR SOLVING DYADIC GREEN''''S FUNCTION OF ELECTROMAGNETIC FIELD

A new method for solving electromagnetic field boundary value problem is given.Byusing this method,the boundary value problem of the vector wave equation can be transformedinto the independent boundary value problem of scalar wave equations and the two additionalvector differential operations.All the dyadic Green's functions got by eigenfunction expansionof the dyadic Green's function can be got by this method easily and some of the dyadic Green'sfunctions for complex systems which are very difficult to get by the ordinary method have beengot by this new method.The dyadic Green's function for a dielectric loaded cavity is one of thegiven examples.     

LOGICAL ANALYSIS FOR NEURAL NETWORKS

By comparison with constraint satisfaction networks, this paper presents an essential frame of the logical theory for continuous-state neural networks, and gives the quantitative analyzing method for contradiction. The analysis indicates that the basic reason for the alternation of the logical states of the neurons is the existence of superior contradiction inside the networks. The dynamic process for a neural network to find a solution corresponds to eliminating the superior contradiction.     

A NEW FRACTAL ZEROTREE CODING FOR WAVELET IMAGE

Based on the mechanisms underlying the performance of fractal and Discrete Wavelet Transform(DWT), one method using fractal-based self-quantization coding way to code different subband coefficients of DWT is presented. Within this method finer coefficients are fractal encoded according to the successive coarser ones. Self-similarities inherent between parent and their children at the same spatial location of the adjacent scales of similar orientation are exploited to predict variation of information across wavelet scales. On the other hand, with respect to Human Visual System(HVS) model, we assign different error thresholds to different decomposition scales, and different shape of range blocks to different orientations of the same scale, by which the perceptually lossless high compression ratio can be achieved and the matching processing can be quickened dramatically.     

A METHOD BASED ON A SIMPLIFIED MODEL FOR DESIGN OF THE GYROTRON GUN

Electron motion in the adiabatic magnetron injection gun(AMIG)of gyrotron is briefly described.A concept of equivalent diode operating with temperature limited emission in the cathode region isproposed.A formula calculating transverse beam energy determined by the electrode dimensionsand electromagnetic parameters is derived from the adiabatic theory and the analysis of electron streamin this diode and the field between two concentric cylinders.With this formula,it is very convenientfor us to rapidly and correctly determine the main parameters of the AMIG or judge the moving con-dition of the electron beam in gyrotron according to practical parameters.The comparison between the calculations of the cathode field in this paper and the numericalresults of the AMIG given by computer shows that the difference is within five percent.The calcula-tions of transverse beam energy are in good agreement with experimental results.     

一种基于子波神经网络的多尺度差分方程求解新方法

多尺度方程的求解是子波与多分辨分析的基础,本文提出了用M-带子波神经网络求解多尺度差分方程(Multiscale Difference Equation, MSDE)的新方法.该方法与以前方法相比,具有很多优点并且求解方程的范围更为广泛.     

基于四进制小波和神经网络的数字水印算法

提出了一种基于BPN神经网络的水印新算法,首先对数字图像进行四进制小波分解;其次,在小波域内选择水印位嵌入的系数块,然后利用神经网络从嵌入水印的图像中恢复原始水印信息。实验结果表明这种水印算法具有很好的性能。     

MSDE信号模型的信号编码方法

本文给出了基于MSDE、LMSDE信号模型的信号编码的数学描述。把LMSDE信号模型的模拟信号编码过程转化为求解一个带非线性约束的二次规划问题并给出了SUMT外点法求解的方法。最后,通过两个例子说明了本文方法的有效性和高信噪比下对噪声的鲁棒性。     

基于小波变换多尺度积的图像融合算法

图像融合是图像处理中的关键技术之一。它在军事和民用图像处理领域获得了广泛的应用。提出了一种新的基于小波变换多尺度积的图像融合算法,小波变换多尺度积具有放大信号边缘特征和降低信号噪声的特点,有利于在融合图像中保持图像的细节特征。利用统计分析的评判准则,如熵、标准偏差评价图像的融合效果。实验结果表明该方法提高了图像的熵和标准偏差。在保留原图像信息的情况下增强了融合图像的细节信息。     

一种新的基于混沌神经网络的动态路由选择算法

针对通信网的路由选择问题,提出了一种动态路由选择的混沌神经网络实现方法。所提出的此方法具有许多优良特性,即暂态混沌特性和平稳收敛特性,能有效地避免传统Hopfield神经网络极易陷入局部极值的缺陷。它通过短暂的倒分叉过程,能很快进入稳定收敛状态。实验证明了本算法能实时、有效地实现通信网的路由选择,并且当通信网中的业务量发生变化时,算法能自动调整最短路径和负载平衡之间的关系。     

基于SOM神经网络的航舵故障分类方法

为解决航舵故障诊断的复杂非线性模式分类问题,提出一种基于自组织特征映射(SOM)神经网络的航舵故障诊断方法,构造一个2层SOM神经网络,训练后多个权值向量位于输入向量聚类中心,实现快速有效的自适应分类.仿真结果表明:SOM网络经过100次训练即可实现聚类,对有限故障测试样本分类准确率可达90％,对航舵故障诊断具有一定的参考价值.     

基于神经网络和支持向量机的电力系统负荷预测方法

提高电力系统负荷预测的精确度是当前负荷预测工作的难点。考虑到神经网络可以逼近任意的非线性关系，而支持向量机能够将约束问题转化，容易地找到全局极小。本文提出了一种基于神经网络和支持向量机的混合负荷预测方法，此方法能通过支持向量机消除了神经网络的总和较小，但单点误差较大的不利现象，而神经网络消除了支持向量机对于模型的简单化问题。最后，负荷预测结果表明本文的方法非常有效。     

基于PSO-RBF神经网络的模拟电路故障诊断研究

针对模拟电路故障诊断的难点和传统诊断方法的不足之处,提出了一种基于PSO算法优化的RBF神经网络模拟电路故障诊断方法。为了约简网络结构从而提高诊断效率,采用主成分分析方法对故障特征进行有效提取。针对RBF网络传统训练算法中隐层节点中心及基函数宽度选取困难问题,提出采用PSO算法来优化训练RBF网络,以提高网络的训练速度和泛化性能。最后,通过电路仿真对所提方法的有效性进行了验证。     

基于组合差异图和卷积小波神经网络的SAR图像变化检测

合成孔径雷达(Synthetic aperture radar,SAR)图像变化检测可以看作是一个分类过程,像素被分类为变化类和不变类.但是,差异图的质量影响了现有方法的检测精度.为了提供高质量的差异图,提出一种基于组合差异图和卷积小波神经网络(Convolutional-Wavelet Neural Network,...