Statistical analysis of state-space approach in harmonic retrieval

The author presents a statistical analysis of the performance of the state-variable balancing for estimating the parameters of exponential signals in the presence of additive noise. The case of frequency estimation for a single damped sinusoid is carried out in detail because the analysis can be presented more clearly; the formulas are simpler and provide insight; and the results are applicable in radar, sonar, communication, and data modelling. Analytical expressions for the variances of the frequency estimate at high signal-to-noise ratios is derived. Both analytical and experimental results are presented to illustrate the performance of the state-variable balancing method     

Cumulant-based approach to harmonic retrieval and related problems

A frequently encountered problem in signal processing is that of estimating the frequencies and amplitudes of harmonics observed in additive colored Gaussian noise. In practice, the observed signals are contaminated with spatially and temporally colored noise of unknown power spectral density. A cumulant-based approach to these problems is proposed. The cumulants of complex processes are defined, and it is shown that specific 1-D slices of the fourth-order cumulant of the noisy signal for the direction of arrival (DOA) and retrieval of harmonics in noise (RHN) problems are identical to the autocorrelation of a related noiseless signal. Hence correlation-based high-resolution methods may be used with fourth-order cumulants as well. The effectiveness of the proposed methods is demonstrated through standard simulation examples     

检测强混沌中微弱谐波信号的神经网络方法

提出了一种提取强混沌中微弱谐波信号的方法。该方法根据嵌入定理，利用混沌系统的单变量观测值对混沌背景重构相空间，并利用径向基函数神经网络(RBFNN)建立混沌噪声的一步预测模型，使其与混沌噪声具有相同的基本动力学特征。并结合一个梳状滤波器对预测误差进行滤波，从而检测出湮没在混沌中的感兴趣的微弱谐波信号。该方法在信噪比(SNR)为-46dB时仍可检测出强混沌中微弱谐波信号。     

粒子群优化RBF神经网络谐波电流检测方法研究

谐波污染问题,众所周知是使用了很多的电力电子器件导致进一步加重,所以谐波治理就变得愈来愈紧要。而瞬时无功功率的检测方法不能满足目前谐波检测对其检测精度的要求,故文中提出一种用粒子群优化RBF神经网络权重的谐波检测方法来检测谐波。改善的方法措施是以瞬时无功功率检测的输入作为RBF神经网络的输入,其输出作为RBF神经网络的期望输出,使用MATLAB软件对RBF神经网络模型进行搭建并进行仿真分析。从实验结果可以得出:使用粒子群优化RBF神经网络权重的谐波检测方法,克服了RBF神经网络检测方法精度不足的欠缺,提高了RBF神经网络的收敛速度和谐波检测的精度。     

Improved Newton-type algorithm for adaptive implementation ofPisarenko's harmonic retrieval method and its convergence analysis

Pisarenko's harmonic retrieval (PHR) method is probably the first eigenstructure based algorithm for estimating the frequencies of sinusoids corrupted by additive white noise. To develop an adaptive implementation of the PHR method, one group of authors has proposed a least-squares type recursive algorithm. In their algorithm, they made approximations for both gradient and Hessian. The authors derive an improved algorithm, where they use exact gradient and a different approximation for the Hessian and analyze its convergence rigorously. Specifically, they provide a proof for the local convergence and detailed arguments supporting the local instability of undesired stationary points. Computer simulations are used to verify the convergence performance of the new algorithm. Its performance is substantially better than that exhibited by its counterpart, especially at low SNR's     

A hybrid approach to harmonic retrieval in non-Gaussian ARMA noise

Addresses the harmonic retrieval problem in colored noise. As contrasted to the reported studies in which Gaussian noise was assumed, this paper focuses on additive non-Gaussian ARMA noise. Our approach is hybrid in the sense that third-order cumulants are first used to identify the AR part of the non-Gaussian noise process, and then correlation-based high-resolution methods are used for the filtered process to estimate the number of harmonics and their frequencies. Simulation examples are presented to demonstrate the high resolution of this approach     

基于主成分分析与BP神经网络的识别方法研究

利用BP神经网络对红外目标进行识别之前,若不对原始样本数据进行预处理与特征提取,一方面使识别结果准确性降低,另一方面使BP神经网络的结构复杂化,采用主成分分析法可解决这些问题。主成分分析法能较好地提取表征样本的少数几个主分量,由该方法的特点可知,这几个主分量彼此不相关,非常符合特征优化的要求。研究结果表明,用该方法处理后的结果数据输入BP神经网络．提高了识别正确率,减少了训练时间,同时也简化了网络结构。将两种常见的模式识别方法结合用于红外目标识别：先由主成分分析法对原始样本数据进行精简处理,然后再由BP神经网络法进行分类识别,与传统的单一识别方法相比,准确度得到提高,计算量大为减少。     

Computational analysis of rotational search methods for adaptivePisarenko harmonic retrieval

The rotational search method (RSM), which reduces the computational burden of the Pisarenko method for identifying undamped sinusoids in additive noise, is discussed. A brief review of previous adaptive algorithms for implementing the method is presented. Two simulations are presented which show the inherent sensitivity of the RSM algorithm. The sensitivity is shown to depend upon the Rayleigh-Ritz eigenvector estimate and the value of the minimum eigenvalue estimate. The errors in their tracking result in a loss of search direction orthogonality to the estimated minimum eigenvector. A higher precision for the enforced search direction orthogonality which minimizes the added computation is proposed, resulting in the altered RSM (ARSM)     

A novel traffic identification approach based on multifractal analysis and combined neural network

An accurate identification of Internet traffic of different applications is highly relevant for a broad range of network management and measurement tasks, including traffic engineering, service differentiation, performance monitoring, and security. Traditional traffic identification approaches have become increasingly inaccurate due to restrictions of port numbers, protocol signatures, traffic encryption, and etc. In this paper, a new traffic identification approach based on multifractal analysis of wavelet energy spectrum and classification of combined neural network models is proposed. The proposed approach is able to achieve the identification of different Internet application traffic by performing classification over the wavelet energy spectrum coefficients that were inferred from the original traffic. Without using any payload information, the proposed approach has more advantages over traditional methods. The experiment results illustrate that the proposed approach has satisfactory identification results.     

基于人工神经网络的图像检索平台的设计与开发

用户描述图像的高层抽象语义与图像内在的底层特征之间存在差异,此时仅依靠图像内容特征进行检索的系统无法准确完成用户的检索任务。针对以上问题,提出了使用神经网络进行图像的匹配计算方法,通过样例自动学习和用户反馈学习两种学习方式,形成图像底层特征到图像分类的正确映射,学习后的神经网络可以进行图像的自动分类及检索。该方法结合了图像的底层特征描述及用户的高层语义反馈,有效地弥补了语义鸿沟。最后,系统通过整合Web前端、图像提取模块、神经网络模块及数据库模块,实现了神经网络学习及图像检索的完整流程。     

Spectroscopy and hybrid neural network analysis

This paper reviews the current use of spectroscopy and related instrumentation in chemical analysis. Advancements in digital signal processing technology are making it possible to improve the sensitivity and accuracy of analytical instruments without expensive upgrading of instrument hardware. A hybrid neural network (HNN) is described that can perform nonlinear signal analysis. The HNN approach combines the simple data reduction capability of conventional linear signal processing algorithms with the adaptive learning and recognition ability of a multilayer nonlinear neural network architecture. A number of examples show the rise of the HNN for environmental monitoring and real-time process control     

Solder joints inspection using a neural network and fuzzyrule-based classification method

In this paper, we described an approach in automation, the visual inspection of solder joint defects of surface mounted components on a printed circuit board, using a neural network with fuzzy rule-based classification method. Inherently, the solder joints have a curved, tiny, and specular reflective surface. This presents the difficulty in taking good images of the solder joints. Furthermore, the shapes of the solder joints tend to greatly vary with their soldering conditions, and are not identical with each other, even though some of the solder joints belong to a set of the same soldering quality. This problem makes it difficult to classify the solder joints according to their properties. To solve this intricate problem, a new classification method is here proposed which consists of two modules: one based upon an unsupervised neural network, and the other based upon a fuzzy set theory. The novel idea of this approach is that a fuzzy rule table reflecting the knowledge of criteria of a human inspector, is utilized in order to correct any possible misclassification made by the neural network module. The performance of the proposed approach was tested on numerous samples of printed circuit boards in commercially available computers, and then compared with that of a human inspector. Experimental results reveal that the proposed method is superior to the neural network classification method alone, in terms of its accuracy of classification     

自适应线性神经网络LMM算法的谐波辨识技术研究

介绍了组合适应线性神经网络最小平均值评估法(Adaline-LMM)对脉冲控制信号的拟合分析方法,用于对电力控制系统中的信号评估。通过对系统信号中的各个谐波分量的幅值和相位进行谐波辨识,并对Adaline的权重向量进行更新,同时对目标函数进行技术估计。其中,自适应神经网络中的权重向量由LMM算法进行迭代更新,通过最小平均值估计算法的引入,减小由于脉冲噪声引起的暂时波动的影响。通过对给定脉冲信号进行拟合,可以发现所提方法具有较高的计算精度。     

Almost-sure identifiability of multidimensional harmonic retrieval

Two-dimensional (2-D) and, more generally, multidimensional harmonic retrieval is of interest in a variety of applications, including transmitter localization and joint time and frequency offset estimation in wireless communications. The associated identifiability problem is key in understanding the fundamental limitations of parametric methods in terms of the number of harmonics that can be resolved for a given sample size. Consider a mixture of 2-D exponentials, each parameterized by amplitude, phase, and decay rate plus frequency in each dimension. Suppose that I equispaced samples are taken along one dimension and, likewise, J along the other dimension. We prove that if the number of exponentials is less than or equal to roughly IJ/4, then, assuming sampling at the Nyquist rate or above, the parameterization is almost surely identifiable. This is significant because the best previously known achievable bound was roughly (I+J)/2. For example, consider I=J=32; our result yields 256 versus 32 identifiable exponentials. We also generalize the result to N dimensions, proving that the number of exponentials that can be resolved is proportional to total sample size     

Two-dimensional harmonic retrieval and its time-domain analysistechnique

The two-dimensional harmonic retrieval is examined in theory by confirming that the 2-D sinusoids in white noise are modeled as a special 2-D autoregressive moving average (ARMA) process whose AR parameters are identical to the MA ones. A new analysis technique for resolving 2-D sinusoids in white noise is proposed     

Design,analysis and application of a volumetric convolutional neural network

The design, analysis and application of a volumetric convolutional neural network (VCNN) are studied in this work. Although many CNNs have been proposed in the literature, their design is empirical. In the design of the VCNN, we propose a feed-forward K-means clustering algorithm to determine the filter number and size at each convolutional layer systematically. For the analysis of the VCNN, the cause of confusing classes in the output of the VCNN is explained by analyzing the relationship between the filter weights (also known as anchor vectors) from the last fully-connected layer to the output. Furthermore, a hierarchical clustering method followed by a random forest classification method is proposed to boost the classification performance among confusing classes. For the application of the VCNN, we examine the 3D shape classification problem and conduct experiments on a popular ModelNet40 dataset. The proposed VCNN offers the state-of-the-art performance among all volume-based CNN methods.     

Development of a neural network based algorithm for radar snowfallestimation

Using radar to measure snowfall accumulation has been a research topic in radar meteorology for decades. Traditionally, a parametric reflectivity-snowfall (Z-S) relationship is used to estimate ground snowfall amounts based on radar observations. However, the accuracy and reliability of Z-S relationship are limited by the wide variability of the Z-S relationship with snowfall type. In this paper, the authors introduce a neural network based approach to address the problem of snowfall estimation from radar by taking into account the vertical structure of precipitation. The motivation for using a multilayer feedforward neural network (MFNN), such as the radial-basis function (RBF) network, is the good universal function approximation capability of the network. The network is trained using vertical reflectivity profiles averaged over a 9-km² area as the input and ground snowfall amounts as the target output. Separate data, which are not part of the training data, are used to test the generalization performance of the RBF network after the training is done. Radar reflectivity data collected by the CSU-CHILL multiparameter radar and ground snowfall measurements recorded by snowgages located at the Stapleton International Airport (SIA), Stapleton, CO, and the Denver International Airport (DIA), Denver, CO, during the Winter and Icing and Storms Projects (WISP94) were used for this study. The snowfall estimates from the RBF network are shown to be better than those obtained from conventional Z-S algorithms. The neural network based approach provides an alternate method to the snowfall estimation problem     

A neural network approach to MVDR beamforming problem

A Hopfield-type neural network approach which leads to an analog circuit for implementing the real-time adaptive antenna array is presented. An optimal array pattern can be steered by updating the weights across the array in order to maximize the output signal-to-noise ratio (SNR). The problem of adjusting the array weights can be characterized as a constrained quadratic nonlinear programming. The adjustment of settings is required to respond to a rapid time-varying environment. A Hopfield-type neural net with a number of graded-response neurons designed to perform the constrained quadratic nonlinear programming would lead to a solution in a time determined by RC time constants, not by algorithmic time complexity. The constrained quadratic programming neural net has associated it with an energy function which the net always seeks to minimize. A fourth-order Runge-Kutta simulation shows that the circuit operates at a much higher speed than conventional techniques and the computation time of solving a linear array of 10 elements is about 0.1 ns for RC=5×10 ^-9