Markov availability models on neural networks

Neural networks are employed to realize an availability model of the TMR hardware with fault-coverage and repair. The failure rate and repair rate of the system that satisfy a target pointwise availability are extracted from the weights of the neural network. The structure and results of the simulation algorithm are presented.     

Microwave diversity imaging using six-port reflectometer

A microwave diversity imaging system conventionally uses a vector network analyzer (VNA) to directly measure the object scattered field (amplitude and phase) over a selected frequency range and viewing angles, then reconstructs the scattering object characteristic function through two dimensional Fourier inversion. In this paper, we present a cost-effective microwave diversity imaging system using a six-port reflectometer, which measures four amplitude (or power) values to acquire the object scattered field indirectly. One can then eliminate the coherent detectors in a VNA. The calibration procedure for this microwave diversity imaging measurement is also described. Experimental results of three types of scattering objects, a metallic cylinder, four distributed line scatterers, and a 72:1 scaled B-52 aircraft model, are presented using the described six-port microwave imaging system     

基于神经网络的GSM无源探测系统目标检测算法

对于GSM相控阵无源雷达接收机获取的目标数据提出一种用最佳后验感知的神经网络进行处理的算法,在复杂的杂波及噪声背景下,相比于流行的卡尔曼滤波,提高了目标的检测跟踪精度,对促进GSM无源探测系统实用化具有重要意义。     

Microwave FinFET modeling based on artificial neural networks including lossy silicon substrate

Nowadays, FinFET represents a new and promising transistor structure for the aggressive downscaling of the CMOS technology. Typically, the small-signal modeling for FinFET is based on compact models or on equivalent circuit representations. As an alternative to such approaches, a small-signal behavioral model based on artificial neural networks is developed in this paper. Particular attention is devoted to modeling the low-frequency kinks of the scattering parameters, due to the lossy silicon substrate. The model is efficient and accurate, as confirmed by the comparison between measured and simulated microwave behavior.     

基于神经网络的机动目标跟踪模糊Kalman滤波算法

以实现坦克对机动目标的有效跟踪为背景,针对传统Kalman滤波算法存在的计算量较大、需要先验信息较多的缺点,提出了一种基于神经网络的机动目标跟踪模糊Kalman滤波算法.在"当前"统计模型的基础上,将未知的目标机动加速度作为附加的过程噪声,使用模糊系统估计全部过程噪声的时变方差,利用神经网络对模糊系统中的参数进行优化.仿真结果表明了所提方法的有效性.     

基于深度谱卷积神经网络的高效视觉目标跟踪算法

提出了一种基于深度频谱卷积神经网络的视觉目标跟踪算法。该算法在深度模型训练阶段采用谱池化替代深度卷积神经网络中的最大池化过程,用贝叶斯分类器替代softmax损失层计算最大分类值,并将其整合到深度神经网络跟踪框架中,通过新网络计算输入正负样本的概率分布预测目标位置。该算法充分利用谱池化在频域下降维到任意维度且计算高效的优点,克服了最大池化采样造成大量空间信息丢失的不足,提升了计算速度。在权威多场景视频标准测试库上对所提算法进行验证,结果验证了该算法兼顾了效率和跟踪精度,有效提高跟踪器的性能,在相同测试条件下,文中算法性能优于同类对比算法。     

On fault probabilities and yield models for VLSI neural networks

We investigate the estimation of fault probabilities and yield for very large scale integration (VLSI) implementations of neural computational models. Our analysis is limited to structures that can be mapped directly onto silicon as truly distributed parallel processing systems. Our work improves on the framework suggested by Feltham and Maly and is also applicable to analog or mixed analog/digital VLSI systems     

Physically realistic fault models for analog CMOS neural networks

A general methodology for the development of physically realistic fault models for VLSI neural networks is presented. The derived fault models are explained and characterized in detail. The application of this methodology to an analog CMOS implementation of fixed-weight (i.e., pretrained), binary-valued neural networks is reported. It is demonstrated that these techniques can be used to accurately evaluate defect sensitivities in VLSI neural network circuitry. It is also shown that this information can be used to guide the design of circuitry which fully utilizes a neural network's potential for defect tolerance     

基于时频图深度学习的雷达动目标检测与分类

雷达动目标检测技术一直是雷达信号处理领域中的关键技术,而传统的雷达动目标检测技术仅适用于匀速运动目标,检测性能有限。针对该问题提出一种基于卷积神经网络(CNN)时频图处理的雷达动目标检测方法,通过从雷达动目标回波中提取多普勒频移信息,然后利用短时傅里叶变换转换为时频图,输入卷积神经网络,进行深度特征学习,进而实现检测和分类的目的。仿真数据验证表明,所提方法能够有效检测和区分匀速、匀变速运动以及微动目标,稳健性高,与传统动目标检测方法相比具有显著优势。     

Millimeter-wave active imaging using neural networks for signal processing

A neural network has been successfully implemented in an active-mode millimeter-wave (60 GHz) imaging system with a Yagi-Uda antenna array in order to recognize objects and reconstruct images that appear distorted under coherent millimeter-wave illumination. With 10 /spl times/ 10 sampling points and five teaching trials, a recognition rate of 98% has been obtained for ten dissimilar alphabetical letters used as objects. The success rate of reconstruction of distorted millimeter-wave images was 80% when five dissimilar letters were used for the reconstruction. The recognition rate after changing the spatial resolution of the optical system and sampling interval of the image is also discussed.     

基于深度卷积神经网络的红外小目标检测

提出了一种新的解决红外图像小目标检测问题的深度卷积网络,将对小目标的检测问题转化为对小目标位置分布的分类问题;检测网络由全卷积网络和分类网络组成,全卷积网络对红外小目标进行增强和初步筛选,实现红外图像的背景抑制,分类网络以原始图像和背景抑制后的图像为输入,对目标点后续筛选,网络中引入SEnet(Squeeze-and-Excitation Networks)对特征图进行选择;实验验证了整个检测网络相对于传统小目标检测算法的优势,所提出的基于深度卷积神经网络的小目标检测方法对复杂背景下低信噪比且存在运动模糊的小目标具有很好的检测效果.     

Estimating one-dimensional models from frequency-domainelectromagnetic data using modular neural networks

An artificial neural network interpretation system is being used to interpret data from a frequency-domain electromagnetic (EM) geophysical system in near real time. The interpretation system integrates 45 separate networks in a data visualization shell. The networks produce interpretations at three different transmitter-receiver (Tx-Rx) separations for half-space and layered-Earth interpretations. Modular neural networks (MNNs) were found to be the only paradigm that could successfully perform the layered-Earth interpretations. An MNN with 16 inputs, five local experts, each with seven hidden processing elements, and three outputs was trained on 4795 patterns for 200 epochs. For two-layer models with a resistivity contrast greater than 2:1, resistivity estimates had greater than 96% accuracy for the first-layer resistivity, greater than 98% for the second-layer resistivity, and greater than 96% for the thickness of the first layer. If the contrast is less than 2:1, the resistivity accuracies are unaffected but thickness estimates for layers less than 2 m are unreliable. A Tx-Rx separation of 16 m with maximum depth of penetration of 8 m was assumed for the example cited     

Using neural networks to construct models of the molecular beamepitaxy process

This paper presents the systematic characterization of the molecular beam epitaxy (MBE) process to quantitatively model the effects of process conditions on film qualities. A five-layer, undoped AlGaAs and InGaAs single quantum well structure grown on a GaAs substrate is designed and fabricated. Six input factors (time and temperature for oxide removal, substrate temperatures for AlGaAs and InGaAs layer growth, beam equivalent pressure of the As source and quantum well interrupt time) are examined by means of a fractional factorial experiment. Defect density, X-ray diffraction, and photoluminescence are characterized by a static response model developed by training back-propagation neural networks. In addition, two novel approaches for characterized reflection high-energy electron diffraction (RHEED) signals used in the real-time monitoring of MBE are developed. In the first technique, principal component analysis is used to reduce the dimensionality of the RHEED data set, and the reduced RHEED data set is used to train neural nets to model the process responses. A second technique uses neural nets to model RHEED intensity signals as time series, and matches specific RHEED patterns to ambient process conditions. In each case, the neural process models exhibit good agreement with experimental results     

Use of neural fuzzy networks with mixed genetic/gradient algorithmin automated vehicle control

This paper is concerned with the design of automated vehicle guidance control. First, we propose to implement the guidance tasks using several individual controllers. Next, a neural fuzzy network (NFN) is used to build these controllers, where the NFN constructs are neural-network-based connectionist models. A two-phase hybrid learning algorithm which combines genetic and gradient algorithms is employed to identify the NFN weightings. Finally, simulations are given to show that the proposed technology can improve the speed of learning convergence and enhance the performance of vehicle control     

Safety-modelling on neural networks

Safety modelling of fault-tolerant hardware is as important as reliability modelling for some applications. Neural networks are employed in this paper to realize the safety model of a Duplex system under design. The failure rate which is adequate for the desired safety of the hardware, is acquired from the neural weights at convergence.     

Microwave detection and dielectric characterization of cylindrical objects from amplitude-only data by means of neural networks

In this paper, we propose a new microwave technique for the localization and the dielectric characterization of physically inaccessible cylindrical objects from amplitude-only data. By means of a neural network used to solve the inverse scattering problem; this technique allows to directly achieve the object retrieval, avoiding the drawbacks related to the measurement of the phase distribution of the field that generally represent a critical point, especially at high frequency. The efficiency of the proposed technique in the reconstruction of both the position and the dielectric properties of a circular cylindrical body from amplitude-only information is illustrated and compared with the reconstruction performances of a neural network imaging technique that makes use of both amplitude and phase of the scattered field. The presence of noisy data is also taken into account, showing the dependence of the reconstruction accuracy on the signal-to-noise-ratio.     

基于深度卷积神经网络的红外过采样扫描图像点目标检测方法

针对红外过采样扫描成像特点,提出一种基于深度卷积神经网络的红外点目标检测方法.首先,设计回归型深度卷积神经网络以抑制扫描图像杂波背景,该网络不含池化层,输出的背景抑制图像尺寸与输入图像一致;其次,对抑制后的图像进行门限检测,提取候选目标小区域原始数据;最后,将候选目标区域数据依次输入分类型深度卷积神经网络以进一步判别目标、剔除虚警.生成大量过采样训练数据有效训练两个深度网络.结果表明,在不同杂波背景下,该方法在目标信杂比增益、检测概率、虚警概率和运算时间等方面,均优于典型红外小目标检测方法,适用于红外过采样扫描系统的点目标检测.     

A quasi-parametric algorithm for synthetic aperture radar target feature extraction and imaging with angle diversity

We present synthetic aperture radar (SAR) target feature extraction and imaging techniques with angle divesity. We first establish a flexible data model that describes each target scatterer as a two-dimensional (2D) complex sequence with arbitrary amplitude and constant phase in range and cross-range. A new algorithm, referred to as the QUasiparametric ALgorithm for target feature Extraction (QUALE), is then presented for SAR target feature extraction via data fusion through angle diversity based on the flexible data model. QUALE first estimates the model parameters, which include, for each scatterer, a 2D arbitrary real-valued amplitude sequence, a constant phase, and scatterer locations in range and cross-reange. QUALE then averages the estimated 2D real-valued amplitude sequence over range by making the assumption that the scatterer radar cross section is approximately consant. QUALE next models the so-obtained 1D sequence with a simple sinc function by assuming that the scatterer is approximately a dihedral (a trihedral is approximated as a very short dihedral) and estimates the relevant sinc function parameters by minimizing a nonlinear least-squares fitting function. Finally, the approximate 2D SAR image is reconstructed by using the estimated features. Numerical examples are given to demonstrate the perfomance of the proposed algorithm.This work was supported in part by AFRL/SNAT, Air Force Research Laboratory, Air Force Materiel Command, USAF, under grant no. F33615-99-1-1507, and the National Science Foundation Grant MIP-9457388. The U.S. Goverment is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright notation thereon.     

Microwave imaging of aircraft

Three methods of imaging aircraft from the ground with microwave radar with quality suitable for aircraft target recognition are described. The imaging methods are based on a self-calibration procedure called adaptive beamforming that compensates for the severe geometric distortion inherent in any imaging system that is large enough to achieve the high angular resolution necessary for two-dimensional target imaging. The signal processing algorithm is described and X-band (3-cm)-wavelength experiments demonstrate its success on commercial aircraft flying into Philadelphia International Airport.<>