Scheduled System Maintenance:
On April 27th, single article purchases and IEEE account management will be unavailable from 2:00 PM - 4:00 PM ET (18:00 - 20:00 UTC).
We apologize for the inconvenience.
By Topic

Neural Network Applications in Electrical Engineering, 2004. NEUREL 2004. 2004 7th Seminar on

Date 23-25 Sept. 2004

Filter Results

Displaying Results 1 - 25 of 91
  • Iterative fuzzy rule base technique for image segmentation

    Publication Year: 2004 , Page(s): 221 - 224
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (638 KB) |  | HTML iconHTML  

    The generic fuzzy rule-based image segmentation algorithm (GFRIS) does not produce good results for images containing non-homogeneous regions, as it does not directly consider texture. A new algorithm is proposed, which includes iterative access to the problem of texture segmentation. A qualitative comparison is made between the segmentation results using this method and the GFRIS algorithm applied to two types of images. The results clearly show that this method exhibits significant improvements over GFRIS, when we have a complex texture structure. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Cross-validation and neural network architecture selection for the classification of intracranial current sources

    Publication Year: 2004 , Page(s): 151 - 158
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (920 KB) |  | HTML iconHTML  

    In the present paper, a new methodological approach, for the classification of first episode schizophrenic patients (FES) against normal controls, is proposed. The first step of the methodology applied is the feature extraction, which is based on the combination of the multivariate autoregressive model with the simulated annealing technique, in order to extract optimum features, in terms of classification rate. The classification, as the second step of the methodology, is implemented by means of an artificial neural network (ANN) trained with the backpropagation algorithm under "leave-one-out cross-validation". The ANN is a multilayer perceptron, the architecture of which is selected after a detailed search. The proposed methodology has been applied for the classification of FES patients and normal controls using as input signals the intracranial current sources obtained by the inversion of event-related potentials (ERP) using an algebraic reconstruction technique. Results implementing the proposed methodology provide classification rates of up to 93%. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural network for finding optimal path in packet-switched network

    Publication Year: 2004 , Page(s): 91 - 96
    Cited by:  Papers (5)  |  Patents (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (684 KB) |  | HTML iconHTML  

    Neural networks are very good candidates for solving different ill-defined problems, due to their high computational speed and the possibility of working with uncertain data. Among others, they represent an efficient tool for solving constrained optimization problems. Under appropriate assumptions, routing in packet-switched networks may be considered as an optimization problem, more precisely, as a shortest-path problem, where the Hopfield type neural network exhibits very good performance. An efficient neural network shortest-path algorithm, inspired by the Hopfield network, is suggested. The routing algorithm suggested is designed to find the shortest path but also it takes into account packet-loss avoidance. The applicability of the proposed model is demonstrated through computer simulations for different full-connected networks with both symmetrical and non-symmetrical links. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A coordination model based control of functional arm manipulation by RBF neural networks

    Publication Year: 2004 , Page(s): 159 - 164
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (718 KB) |  | HTML iconHTML  

    A model based control system for neuro-rehabilitation of the upper arm in post-stroke hemiplegic patients was developed. The control system was based on normal values of motion parameters of 6 daily task activities. Kinematic data (6 arm joint angles) was measured by using gonio and torsiometers. From computed angular velocities, the following sequences were extracted: reaching & grasping, manipulation, releasing, and returning hand to resting position. The angular accelerations were calculated in order to create synergies in the form of phase plots used to train radial basis function (RBF) neural networks. The networks generated automatic synergy recognition and classification of arm movements in regard to two workspace attributes: distance and laterality of the object position. The synergies have been used in order to shift the control of multijoint arm movements to a higher level and minimize the number of unique couplings between joint accelerations, which define the task, position, or their combination. One task, eating finger food, was selected to illustrate the methodology as an example of precision grasp. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Alternative signal detection for neural network-based smart antenna

    Publication Year: 2004 , Page(s): 85 - 89
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (612 KB) |  | HTML iconHTML  

    Neural network-based smart antennas are used for the solution of multiple-source tracking problems in the area of wireless communications. The architecture of the neural network is constructed in two stages, one stage for signal detection and the other for angle of arrival (AOA) estimation. The best candidates for this type of problem are radial basis function neural networks (RBFNN), applied in both stages. Progress is made by applying probabilistic neural networks (PNN) in the first stage. This rapidly reduces the time for network training. Simulation results are performed to investigate the performance of the algorithm. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The hybrid-neural empirical model for the electromagnetic field level prediction in urban environments

    Publication Year: 2004 , Page(s): 189 - 192
    Cited by:  Papers (4)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (594 KB) |  | HTML iconHTML  

    The application of multilayer perceptron networks to calculating the electromagnetic wave path loss in an urban environment for propagation through an area with low or high buildings is presented. A hybrid neural-empirical model, created in two phases, is proposed. The first phase implies the realization of an approximate (coarse) propagation model based on measured values. This model determines the propagation loss from the beginning of the area, based on the distance from the area beginning, the average building density, the partial loss of a single building, the distance from the transmitter and the exponential loss index of the area. In the second phase, a neural network and the approximate model are integrated in the hybrid (fine) model of the propagation area. The input parameters for the neural network are the distance from the area beginning and the average height of buildings in that area, while the output parameter is the partial loss of a single building. This value is used in the approximate model, in order to obtain the propagation area model with higher accuracy. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Control of heating, ventilation and air conditioning system based on neural network

    Publication Year: 2004
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (433 KB)  

    Summary form only given. The HVAC (heating ventilation and air conditioning) car unit may be considered as a very complex multiple-input multiple-output system with the basic target to make the car passengers feel comfortable. The system inputs are: air temperature, actual temperature in the passenger compartment, sun load, vehicle speed, desired temperature set by the driver or co-driver, air recirculation request and defrost request. The last two inputs are binary signals. The system outputs are defined by the actuators located in the HVAC unit: defrost door position (it controls the air distribution between the defrost registers and penal-feet registers), mode door position (it controls the air distribution between the penal and feet registers), blend door position that mixtures hot and cold air, recirculation door position that makes a difference between the fresh air input and recirculation and the last output is the blower speed. The system that has to be controlled is quite nonlinear and nonstationary and there are two important criteria that have to be fulfilled. The first of them is the passengers' comfort and this criterion is vague, hardly defined numerically. In order to acquire some training sets many test trips have been made. Car companies usually organize three test trips trying to cover extreme temperature and weather environments. One of them is during summer choosing some hot destinations in Spain, Portugal or North Africa. The second trip is during the winters with the destinations in North Europe and the third trip is the so-called 'mild' trip usually organized in Middle Europe. The set of characteristic training values that may be collected during these trips is usually not enough for training of neural networks. This paper presents an algorithm to design the simulator in order to generate enough data necessary to train the networks. These data must be consistent with the data collected during the trips. Also, the paper proposes an architecture and a training algorithm for the neural networks that control the actuators in the HVAC unit. The obtained results demonstrate the accuracy and the simplicity of the proposed solutions. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A time decoding realization with a CNN

    Publication Year: 2004 , Page(s): 97 - 102
    Cited by:  Papers (25)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (702 KB) |  | HTML iconHTML  

    Time encoding is a novel real-time asynchronous mechanism for encoding amplitude information into a time sequence. The analog bandlimited input is fed into a simple nonlinear neuron-like circuit that generates a strictly increasing time sequence based on which the signal can be reconstructed. The heart of the reconstruction is solving a system of ill-conditioned linear equations. The paper shows that the equations can be manipulated so that the reconstruction becomes feasible using a cellular neural network (CNN) with a banded system matrix. In particular, the system is first transformed into a smaller well-conditioned system, and, then, the Lanczos process is used to lay it out into a set of even smaller systems characterized by a set of tridiagonal matrices. Each of these systems can directly be solved by CNNs, whereas the preprocessing (transformation and Lanczos algorithm) and simple postprocessing phases can be partly or fully implemented by using the digital capabilities of the CNN universal machine (CNN-UM). Each step of the proposed formulation is confirmed by numerical (digital) simulations. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Robust speech recognizer using multiclass SVM

    Publication Year: 2004 , Page(s): 63 - 66
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (602 KB) |  | HTML iconHTML  

    In this paper a robust speech recognizer is presented based on features obtained from the speech signal and also from the image of the speaker. The features were combined by simple concatenation, resulting in composed feature vectors to train the models corresponding to each class. For recognition, the classification process relies on a very effective algorithm, namely the multiclass SVM. Under additive noise conditions the bimodal system based on combined features acts better than the unimodal system, based only on the speech features, the added information obtained from the image playing an important role in robustness improvement. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Fuzzy Petri net based reasoning for the diagnosis of bus condition

    Publication Year: 2004 , Page(s): 225 - 229
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (679 KB) |  | HTML iconHTML  

    A bus maintenance system in a transport organization is discussed. The bus fleet consists of different types of buses operating on international, intercity and local lines. The fleet should have good corrective and preventive maintenance. Periodic inspections and preventive maintenance activities must be conducted following the manufacturer's recommendations for the warranty to be valid and for the equipment to operate properly. Daily requests for maintenance resources vary depending on stochastic factors, which are out of the control of the operations manager. The paper presents a programme developed by the use of a fuzzy Petri net which is aimed to help a maintenance manager to predict the needs for maintenance resources, as well as the time duration of maintenance activities. The prediction is based on the use of incomplete and vague information on the overall condition of a bus. By using CPN Tools software, a colored Petri net that performs the process of fuzzy reasoning is developed. The final assessment of the bus overall condition is drawn from the data and information about the bus and its exploitation in the previous period. The reasoning process consists of five phases, twelve attributes and 45 rules. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Design methods for CNN spatial filters with circular symmetry

    Publication Year: 2004 , Page(s): 103 - 108
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (702 KB) |  | HTML iconHTML  

    We propose an efficient method for designing circularly-symmetric spatial linear filters implemented on cellular neural networks. The proposed method is relatively simple, relying on a 1D prototype filter, for which the templates are determined, and on a spatial frequency transformation. Some design examples are given, for 2D low-pass and band-pass filters (both of FIR and IIR type) with imposed cut-off or peak frequency and a specified selectivity. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Crack width prediction of reinforced concrete structures by artificial neural networks

    Publication Year: 2004 , Page(s): 39 - 44
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (640 KB) |  | HTML iconHTML  

    This paper proposes the use of artificial neural networks (ANN) for the prediction of the maximum surface crack width of precast reinforced concrete beams joined by steel coupler connectors and anchor bars (jointed beams). Two different training algorithms are used in this study and their performance is compared. The first approach used backpropagation and the second one includes genetic algorithms during the training process. Input and output vectors are designed on the basis of empirical equations available in the literature to estimate crack widths in common reinforced concrete (RC) structures and parameters that characterize the mechanical behavior of RC beams with overlapped reinforcement. Two well-defined points of loading are considered in this study to demonstrate the suitability of this approach in both a linear and a highly nonlinear stage of the mechanical response of this type of structure. Remarkable results were obtained, however, in all cases using the combined genetic artificial neural network (GANN) approach which resulted in improved prediction performance over networks trained by error backpropagation. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Power transformer differential protection scheme based on symmetrical component and artificial neural network

    Publication Year: 2004 , Page(s): 261 - 265
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (608 KB) |  | HTML iconHTML  

    This paper proposes a differential protection scheme for power transformers using the symmetrical component and neural network algorithms. It utilizes the artificial neural network (ANN) as the pattern classifier and the symmetrical component of current as the input's ANN. Extensive simulation studies show that the symmetrical components of current provide suitable inputs for classification or different transient cases. The proposed scheme achieves outstanding performance and the ability to discriminate internal faults fast and accurately. Details of the proposed relay design are given in the paper. Some performance studies results are also given. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A MEMS based architecture of artificial neuron

    Publication Year: 2004 , Page(s): 113 - 116
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (597 KB) |  | HTML iconHTML  

    In this paper we propose an optical artificial neuron with synapses fulfilled through reflected light from MEMS mirrors. Its structure is thoroughly described and a detailed critical analysis of neuron performance is given, including issues of number of inputs, the change of synaptic weight, synapse weight setting precision to the prescribed value and even the influence of acoustic noise. Then we draw a conclusion about the effectiveness of such a configuration showing the applicability of MEMS technology to artificial neural networks, and give proposals for attaining effective neural systems. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A fuzzy system for municipalities classification

    Publication Year: 2004 , Page(s): 231 - 236
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (671 KB) |  | HTML iconHTML  

    A software system for municipality classification based on fuzzy rules is presented. The classification is made with the purpose of estimating and roughly planning the road maintenance needs in winter. The knowledge-based system is developed using Matlab and GIS technology. Mamdani and Sugeno type models for fuzzy reasoning have been implemented. Real data from an urban information system used in the Statistical Office of the Slovak Republic are used and knowledge acquisition from experts has been performed. The application of the system is illustrated by the example of classifying municipalities in one of the eight regions in the Slovak Republic. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural networks in microwave low-noise transistor modeling under various temperature conditions

    Publication Year: 2004 , Page(s): 199 - 203
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (664 KB) |  | HTML iconHTML  

    The paper considers neural network based modeling of the temperature dependence of low-noise microwave transistor signal and the noise performance. Two approaches are proposed. One is a combination of a transistor empirical model and neural networks and the other is completely based on neural networks. Implementation of the proposed models into a microwave simulator is described. For a specified component, comparison results are given of signal and noise parameter prediction using these models under various temperature conditions. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Specific process models derived from extremely small data sets and general process models

    Publication Year: 2004 , Page(s): 267 - 272
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (682 KB) |  | HTML iconHTML  

    Definition of a needed particular process model is based on a combination of weighted known general process models and standard error minimization. The known general process models correspond to the biological processes of growing. The standard error is computed using new data and an ensemble of generated models. General models are based on polynomial functions and neural networks. Applications of polynomial functions of second, third and fourth degrees is analyzed. Supervised learning of the neural networks is based on the Levenberg-Marquardt algorithm. A very brief comment on the Vapnik-Chervonenkis dimension as an important parameter in modern learning theory, is also done in view of the analyzed cases. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Increasing the efficiency of radiolocation systems with application of artificial intelligence

    Publication Year: 2004 , Page(s): 139 - 143
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (589 KB) |  | HTML iconHTML  

    The specialized radiolocation complexes working with spread receivers provide high obscurity, noise protection and reliability. These demands can be fulfilled with complex signals emitted from the source, radioholographic approaches in processing and a decision making method for identification. In the paper the application of artificial intelligence for increasing the efficiency of radiolocation systems (RLS) is introduced. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Application of artificial intelligence techniques to obtain robust dynamic equivalents

    Publication Year: 2004 , Page(s): 51 - 54
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (611 KB) |  | HTML iconHTML  

    This paper presents an application of artificial neural networks (ANN) to power systems. ANN are tested to construct dynamic equivalents, which is considered a hard task in the context of power systems. The main objective is to reproduce the complex voltage at frontier nodes. The simulation results prove the applicability and robustness of this innovative approach. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Modified ANFIS architecture - improving efficiency of ANFIS technique

    Publication Year: 2004 , Page(s): 215 - 220
    Cited by:  Papers (4)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (685 KB) |  | HTML iconHTML  

    Adaptive neuro-fuzzy inference systems (ANFIS), fusing the capabilities of artificial neural networks and fuzzy inference systems, offer a lot of space for solving different kinds of problems, and are especially efficient in the domain of signal prediction. However, the ANFIS technique is sometimes notated as being computationally expensive. The paper, after considering the conventional ANFIS architecture, brings up a modified ANFIS (MANFIS) structure developed with the intention of making the ANFIS technique more efficient with regard to root mean square error (RMSE) and/or computing time. The standard benchmark, prediction of the Mackey-Glass time series, was used to prove the better performance of the proposed MANFIS structure. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Does iterative nonlinear neural adaptive filtering affect the nature of the processed signal?

    Publication Year: 2004 , Page(s): 21 - 24
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (598 KB) |  | HTML iconHTML  

    The data-reusing (DR) approach is commonly used to improve the convergence rate and robustness of standard adaptive filters. However, it is largely unknown whether such an approach affects the linear/nonlinear nature of the processed signal. It is therefore natural to ask ourselves a question "does iterative nonlinear neural adaptive filtering affect the nature of the processed signal". To help to answer this, we provide a quality assessment of both the standard and data-reusing direct gradient algorithms for linear FIR filters and neural networks applied in adaptive filtering. This is achieved based upon some recently introduced phase space based methods for signal characterisation. A comprehensive analysis on both linear and nonlinear benchmark signals suggests that data-reusing algorithms not only exhibit a performance advantage over the standard algorithms, but also that the processed signal nature matching improves with the order of DR iteration. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • On the recurrent neural networks for solving general quadratic programming problems

    Publication Year: 2004 , Page(s): 5 - 9
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (579 KB) |  | HTML iconHTML  

    Quadratic programming problems are a widespread class of nonlinear programming problems with many practical applications. The case of inequality constraints have been considered in a previous author's paper. In this contribution, an extension of these results for the case of inequality and equality constraints is presented. Based on an equivalent formulation of the Kuhn-Tucker conditions, a new neural network for solving general quadratic programming problems, for the case of both inequality and equality constraints, is proposed. Two theorems for global stability and convergence of this network are given as well. The presented network has lower complexity for implementations and the examples confirm its effectiveness. Simulation results based on SIMULINK® models are given and compared. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Synergy of classical and quantum communications channels in brain: neuron-astrocyte network

    Publication Year: 2004 , Page(s): 177 - 182
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (660 KB) |  | HTML iconHTML  

    The human brain acts as a very efficient parallel classical-quantum communications system relating responses to a multitude of input signals from sensors. However, the mechanisms of the quantum communication channel in the brain remain a mystery. Since matter and energy cannot be teleported, in the sense of transferring from one place to another without passing through intermediate locations, teleportation of quantum states, as the ultimate structure of objects, is possible. Only the structure, information about conformation state, is teleported - the matter stays at the source side and has to be already present at the final location. From this point of view, we consider a new type of network in the brain, based on a unity of classical and quantum neural-astrocyte networks. Matter and energy (classical channel) and classical quantum information travel through neurons, while through the astrocyte-neuron loop travel classical quantum and non-classical quantum information. The new type of network is a synergetic one with quite new energy-information properties compared to classical neuron networks. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Mechanism for automated neural network based transport system with learning

    Publication Year: 2004 , Page(s): 243 - 248
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (667 KB) |  | HTML iconHTML  

    The wide applicability of neural networks to solve various real life problems is well understood. This paper proposes a mechanism for automated neural network based transport system with learning (MANTLE), an automated transport system with considerable advantages over previous attempts. The system uses a multilayer feed-forward neural network with back propagation learning. In addition, the design of MANTLE involves the convergence of a plethora of technologies like Global Positioning System (GPS), a geographic information system (GIS), and laser ranging. MANTLE can guide a mobile agent through a hostile and unfamiliar domain after being trained by a human user with domain expertise. One of the many areas in which MANTLE scores against the competition is that the system is completely domain independent and incurs substantially less processor overhead. MANTLE thus provides more functionality, even though it requires a lot less input as compared to other attempts in this field. This reduction in the size of the input vector translates into more efficient and faster processing. Another of MANTLE's hallmark features is its ability to negotiate turns and implement lane-changing maneuvers with a view to overtaking obstacles. It does this by employing a novel technique, selective net masking. A simulation of MANTLE's neural network was performed on a variety of network topologies, and the best network selected. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Automatic classification of underwater sonar signals

    Publication Year: 2004 , Page(s): 121 - 125
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (586 KB) |  | HTML iconHTML  

    Much work has been performed recently in the area of automatic recognition of the sonar signals to reduce the operator load when confronted with many beams of data concurrently. In this paper we applied and compared different feature extraction methods such as the wavelet basis method, discrete cosine transform method (DCT) and linear prediction method (LP) and then we applied neural network techniques to the processing of sonar signal classification. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.