ACES Publication Search





There are 9 search results for:



Title: July 2003 Full Issue
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 3579 KB

Title: Front / Back Matter and Index
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 795 KB

Title: A Comparative Study of NN and SVM-Based Electromagnetic Inverse Scattering Approaches to On-Line Detection of Buried Objects
Abstract: Microwave-based sensing techniques constitute an important tool for the detection of buried targets. In this framework, a key issue is represented by real-time scatterer localization. As far as such a topic is concerned, this paper presents a comparative evaluation of the performances provided by a conventional NN-based inverse scattering technique and by a new SVM-based electromagnetic approach. In order to estimate the effectiveness values of the two methods, realistic configurations and noisy enviornments are considered and current capabilities, as well as potential limitations, are pointed out. Finally, possible future research work is outlined.
Author(s): Salvatore Caorsi, Davide Anguita, Emanuela Bermani, Andrea Boni, Massimo Donelli
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 319 KB

Title: Neural Network Approaches To The Processing of Experimental Electro-Myographic Data from Non-Invasive Sensors
Abstract: Learning theories and algorithms for both supervised and unsupervised Neural Networks (NNs) have already been accepted as relevant tools to cope with difficult problems based on the processing of experimental electromagnetic data. These kinds of problems are typically formulated as inverse problems. In this paper, in particular, the electrical signals under investigations derive from experimental electromyogram interference patterns measured on human subjects by means of non-invasive sensors (surface ElectroMyoGraphic, sEMG, data). The monitoring and the analysis of dynamic sEMG data reveals important information on muscles activity and can be used to clinicians for both preventing dramatic illness evolution and improving athletes performance. The paper proposes the use of Independent Component Analysis (ICA), an unsupervised learning technique, in order to process raw sEMG data by reducing the typical “cross-talk” effect on the electric interference pattern measured by the surface sensors. The ICA is implemented by means of a multi-layer NN scheme. Since the IC extraction is based on the assumption of stationarity of the involved sEMG recording, which is often inappropriate in the case of biomedical data, we also propose a technique for dealing with non-stationary recordings. The basic tool is the wavelet (time-frequency) decomposition, that allows us to detect and analyse time-varying signals. An auto-associative NN that exploits wavelet coefficients as an input vector is also used as simple detector of non-stationarity based on a measure of reconstruction error. The proposed approach not only yields encouraging results to the problem at hand, but suggests a general approach to solve similar relevant problems in some other experimental applications of electromagnetics.
Author(s): Francesco Carlo Morabito, Maurizio Campolo
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 908 KB

Title: A Combined State Space Formulation/Equivalent Circuit and Neural Network Technique for Modeling of Embedded Passives in Multilayer Printed Circuits
Abstract: : In this paper, we present a new approach for modeling the high-frequency effects of embedded passives in multilayer printed circuits, utilizing state space equations or equivalent circuit together with neural network techniques. In this approach, the neural network based model structure is trained using full wave electromagnetic (EM) data. The resulting embedded passive models are accurate and fast, can be used in both frequency/time domain simulators. Examples of embedded resistor and capacitor models emonstrate that the combined model can accurately represent EM behavior in microwave/RF circuit design. In high-level circuit design, we applied our combined EM based neural models for signal integrity analysis and design of multilayer circuit to illustrate that the geometrical parameters can be continuously adjusted by using neural network techniques. Optimization and Monte-Carlo analysis are performed showing that the combined models can be efficiently used in place of computationally intensive EM models of embedded passives to speed up circuit design.
Author(s): X. Ding, J.J. Xu, M.C.E. Yagoub , Q.J. Zhang
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 438 KB

Title: One-vs-One Multiclass Least Squares Support Vector Machines for Direction of Arrival Estimation
Abstract: This paper presents a multiclass, multilabel implementation of Least Squares Support Vector Machines (LSSVM) for DOA estimation in a CDMA system. For any estimation or classification system the algorithm’s capabilities and performance must be evaluated. This paper includes a vast ensemble of data supporting the machine learning based DOA estimation algorithm. Accurate performance characterization of the algorithm is required to justify the results and prove that multiclass machine learning methods can be successfully applied to wireless communication problems. The learning algorithm presented in this paper includes steps for generating statistics on the multiclass evaluation path. The error statistics provide a confidence level of the classification accuracy.
Author(s): Judd A. Rohwer, Chaouki T. Abdallah
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 453 KB

Title: NEURAL NETWORKS FOR THE CALCULATION OF BANDWIDTH OF RECTANGULAR MICROSTRIP ANTENNAS
Abstract: Neural models for calculating the bandwidth of electrically thin and thick rectangular microstrip antennas, based on the multilayered perceptrons and the radial basis function networks, are presented. Thirteen learning algorithms, the conjugate gradient of Fletcher-Reeves, Levenberg-Marquardt, scaled conjugate gradient, resilient backpropagation, conjugate gradient of Powell-Beale, conjugate gradient of Polak-Ribiére, bayesian regularization, one-step secant, backpropagation with adaptive learning rate, Broyden-Fletcher-Goldfarb-Shanno, backpropagation with momentum, directed random search and genetic algorithm, are used to train the multilayered perceptrons. The radial basis function network is trained by the extended delta-bar-delta algorithm. The bandwidth results obtained by using neural models are in very good agreement with the experimental results available in the literature. When the performances of neural models are compared with each other, the best results for training and test were obtained from the multilayered perceptrons trained by the conjugate gradient of Powell-Beale and Broyden-Fletcher-Goldfarb-Shanno algorithms, respectively.
Author(s): S. Sinan Gultekin, Kerim Guney, Seref Sagiroglu
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 207 KB

Title: APPLICATION OF NEURAL NETWORKS IN THE ESTIMATION OF TWO-DIMENSIONAL TARGET ORIENTATION
Abstract: A new method for the robust estimation of target orientation using measured radar cross section is proposed. The method is based on a Generalized Regression Neural Network (GRNN) scheme. The network is trained by the FFT modulus of bistatic radar cross section data sampled at the receiver positions. The target value to be trained is the angle between a defined target orientation and the incident wave. Results based on actual measurements are presented
Author(s): A. Kabiri, N. Sarshar, K. Barkeshli
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 257 KB

Title: Application of Two-Dimensional AWE Algorithm in Training Multi-Dimensional Neural Network Model
Abstract: Artificial neural network (ANN) plays very important role in microwave engineering. Training a neural network model is the key of neural network technique. The conventional methods for training, such as method of moment (MoM), are time-consuming when the training parameters are a bit more. In order to aid the training process by reducing the amount of costly and time-consuming sampling cycles, a lot of algorithms have been developed, such as asymptotic waveform evaluation (AWE). In this paper, MoM in conjunction with the two-dimensional AWE is applied to accelerate the process of training the neural network model based on the input impedance response on frequency and that on other parameters of a microstrip antenna. In AWE method, the derivatives of Green’s function are required. A closed form of microstrip Green’s function is used for this requirement. Then, the derivative matrices respect to both frequency and permittivity can be obtained from the original matrix. With these matrices in hand, coefficients of the two-dimensional Padé polynomial can be obtained. So the sampling data for training neural network model can be obtained and the process of training neural net model can be completed quickly and accurately. Numerical results demonstrate the efficiency of this technique.
Author(s): Y. Xiong, D. G. Fang, R. S. Chen
File Type: Journal Paper
Issue:Volume: 18      Number: 2      Year: 2003
Download Link:Click here to download PDF     File Size: 324 KB