Search

A framework for the exploitation of a general non-redundant precoding for blind channel estimation in multiple-input-muliple-output (MIMO) block transmission systems with a cyclic prefix (CP) is presented. Findings reveal the achievement of accurate channel estimation with a small number of symbol blocks.

Distributed estimation of deterministic vector parameters using ad hoc wireless sensor networks (WSNs) is studied. The resultant algorithm can be decomposed into a set of simpler tasks suitable for distributed implementation.

The support vector regression framework is proposed as the basis for a new antenna array characterization technique able to model radiating structures including all the coupling effects between the elements of the array from experimental data. The obtained model can then be used to solve practical synthesis problems with high accuracy and using an appropriate cost function for template-based specification handling. A design example is presented to show the efficiency of the proposed method. Index Terms--Antenna arrays, antenna pattern synthesis, support vector regression.

This study explains how a Steinmetz compensator works. Past explanations are based on phasor diagrams and do not address the basic physical phenomena of electromagnetic energy flow. This paper uses the Poynting vector to detect and understand the process of compensation and restoration of symmetry in an unbalanced system. Index Terms--Compensation, power flow, Poynting vector (PV), three-phase systems.

This paper presents the application of a space vector modulation (SVM) strategy for a multimodule converter system that 1) enables low switching frequency, 2) eliminates/minimizes ac-side voltage harmonics, particularly low-order harmonics, and 3) provides maximum ac-side fundamental voltage component. The SVM strategy is based on a sequential sampling technique. The SVM provides harmonic cancellation/minimization by introducing appropriate phase shift for the corresponding voltage harmonics of the converter modules, while maintaining the fundamental voltage components of modules in-phase to obtain maximum ac-side voltage. The SVM eliminates the need for complicated transformer arrangement for harmonic reduction, and thus provides high degree of modularity by utilization of identical transformers for converter modules. The proposed SVM strategy is developed for a back-to-back HVDC converter station in which each converter system is composed of four identical modules. Based on a dynamic model of the system, converter controls are designed and performance of the SVM strategy in terms of converter harmonics and dynamic performance are presented. The studies are performed in time-domain, using the PSCAD/EMTDC software tool. Index Terms--Dynamic model, harmonics, HVDC, multimodule converter, space vector modulation (SVM).

A new image fusion method by employing the support value transform which used the support value to represent the salient features of image based on the support vector machines (SVMs) is described.

The Green's functions for the magnetic scalar and electric vector potentials are derived for a rectangular waveguide with dielectric-filled corrugations supporting left-hand, as well as right-hand propagation. The derivation of the Green's functions for this pair of auxiliary potentials is more involved than their electric-type counterparts. An investigation of the divergence of the electric vector potential is performed with emphasis on the waveguide transverse electric modes. As a result of the divergenceless nature of the vector potential for these modes, the scalar potential is shown to vanish within the corrugations, and the rest of the derivation proceeds by employing the spectral representation then transforming back to the spatial domain. The derived expressions are verified by considering some examples involving discontinuities that can be modeled by equivalent magnetic currents in view of the equivalence principle. Comparison with full-wave simulation commercial packages validates the current theory. Index Terms--Auxiliary potentials, corrugated waveguide, Green's function, metamaterial transmission line.

In this paper, we propose a novel algorithm for rule extraction from support vector machines (SVMs), termed SQRex-SVM. The proposed method extracts rules directly from the support vectors (SVs) of a trained SVM using a modified sequential covering algorithm. Rules are generated based on an ordered search of the most discriminative features, as measured by interclass separation. Rule performance is then evaluated using measured rates of true and false positives and the area under the receiver operating characteristic (ROC) curve (AUC). Results are presented on a number of commonly used data sets that show the rules produced by SQRex-SVM exhibit both improved generalization performance and smaller more comprehensible rule sets compared to both other SVM rule extraction techniques and direct rule learning techniques. Index Terms--Artificial Intelligence, machine learning, representations, information extraction, pattern recognition applications, SVMs.

Support vector machines (SVMs) usually need a large number of support vectors to form their output. Recently, several models have been proposed to build SVMs with a small number of basis functions, maintaining the property that their hidden-layer weights are a subset of the data (the support vectors). This property is also present in some algorithms for feedforward neural networks (FNNs) that construct the network sequentially, leading to sparse models where the number of hidden units can be explicitly controlled. An experimental study on several benchmark data sets, comparing SVMs and the aforementioned sequential FNNs, was carried out. The experiments were performed in the same conditions for all the models, and they can be seen as a comparison of SVMs and FNNs when both models are restricted to use similar hidden-layer weights. Accuracies were found to be very similar. Regarding the number of support vectors, sequential FNNs constructed models with less hidden units than standard SVMs and in the same range as 'sparse' SVMs. Computational times were lower for SVMs. Index Terms--Feedforward neural networks (FNNs), sparse models, support vector machines (SVMs).

The problem of detecting a signal is considered by using multiple observations in the primary data and under a possible scaling inhomogeneity between the samples under test and the training samples. The new detector has provided additional robustness when compared to a detector that assumes a known signature.

A semantic categorization method for generic home photo is proposed. The main contribution of this paper is to exploit a two-layered classification model incorporating camera metadata with low-level features for multilabel detection. The two-layered support vector machine (SVM) classifiers operate to detect local and global photo semantics in a feed-forward way. The first layer aims to predict likelihood of predefined local photo semantics based on camera metadata and regional low-level visual features. In the second layer, one or more global photo semantics is detected based on the likelihood. To construct classifiers producing a posterior probability, we use a parametric model to fit the output of SVM classifiers to posterior probability. A concept merging process based on a set of semantic-confidence maps is also presented to cope with selecting more likelihood photo semantics on spatially overlapping local regions. Experiment was performed with 3086 photos that come from MPEG-7 visual core experiment two official databases. Results showed that the proposed method would much better capture multiple semantic meanings of home photos, compared to other similar technologies. Index Terms--Camera metadata, image classification, photo album, support vector machine.

A new subspace direction-of-arrival (DOA) estimation procedure is developed by using a noneigenvector basis and the computation is carried out by a modified version of the orthogonal auxiliary-vector (AV) filtering algorithm. The new AV basis is used for the DOA estimation with remarkable success relative to (grid-)maximum-likelihood (ML)- type estimation for uncorrelated or even highly correlated sources.

A Lloyd-type vector quantization (VQ) design algorithm and scalar quantization (SQ) are developed for quantizing the phase angles and the performance of quantized equal gain transmission (Q-EGT) is analyzed. The results have shown that although both VQ and SQ achieve the same rate of convergence as the number of feedback bits B increases, a fixed gap is present between the two.

This paper presents a new classifier called total margin-based adaptive fuzzy support vector machines (TAF-SVM) that deals with several problems that may occur in support vector machines (SVMs) when applied to the face recognition. The proposed TAF-SVM not only solves the overfitting problem resulted from the outlier with the approach of fuzzification of the penalty, but also corrects the skew of the optimal separating hyperplane due to the very imbalanced data sets by using different cost algorithm. In addition, by introducing the total margin algorithm to replace the conventional soft margin algorithm, a lower generalization error bound can be obtained. Those three functions are embodied into the traditional SVM so that the TAF-SVM is proposed and reformulated in both linear and nonlinear cases. By using two databases, the Chung Yuan Christian University (CYCU) multiview and the facial recognition technology (FERET) face databases, and using the kernel Fisher's discriminant analysis (KFDA) algorithm to extract discriminating face features, experimental results show that the proposed TAF-SVM is superior to SVM in terms of the face-recognition accuracy. The results also indicate that the proposed TAF-SVM can achieve smaller error variances than SVM over a number of tests such that better recognition stability can be obtained. Index Terms--Face recognition, kernel Fisher's discriminant analysis (KFDA), support vector machines (SVMs).

Many researchers have studied reversible data hiding techniques in recent years and most have proposed reversible data hiding schemes that guarantee only that the original cover image can be reconstructed completely. Once the secret data are embedded in the compression domain and the receiver wants to store the cover image in a compression mode to save storage space, the receiver must extract the secret data, reconstruct the cover image, and compress the cover image again to generate compression codes. In this paper, we present a reversible data hiding scheme based on side match vector quantization (SMVQ) for digitally compressed images. With this scheme, the receiver only performs two steps to achieve the same goal: extract the secret data and reconstruct the original SMVQ compression codes. In terms of the size of the secret data, the visual quality, and the compression rate, experimental results show that the performance of our proposed scheme is better than those of other information hiding schemes for VQ-based and SMVQ-based compressed images. The experimental results further confirm the effectiveness and reversibility of the proposed scheme. Index Terms--Reversible data hiding, side match vector quantization (SMVQ), steganography.

This paper describes a modification of the vector fitting (VF) procedure for rational function approximation of frequency-domain responses. The modification greatly improves the ability of VF to relocate poles to better positions, thereby improving its convergence performance and reducing the importance of the initial pole set specification. This is achieved by replacing the high-frequency asymptotic requirement of the VF scaling function with a more relaxed condition. Calculated results demonstrate a major improvement of performance when fitting responses that are contaminated with noise. The procedure is also shown to be advantageous for wideband modeling of transmission lines, network equivalents, and transformers. Index Terms--Macromodei, rational approximation, system identification, vector fitting (VF).

Accurate diagnosis of faults in transformers can significantly enhance the safety, reliability, and economics of power systems. In the case of a fault, it has been established that the pattern of the fault currents contain a typical signature of the nature and location of the fault for a given winding. This paper describes a new approach using wavelet transform (WT) for extraction of features from the impulse test response of a transformer in time-frequency domain and support vector machine in regression mode to classify the patterns inherent in the features extracted through the WT of different fault currents. This paper also describes an approach to identify the type and location of the transformer faults accurately by analyzing experimental impulse responses that contain noise. Here, experimental impulse responses have been pre-processed with the help of wavelet-packet filters to remove the unwanted noise from the signal and thereby enhance the analyzing capability of continuous wavelet transform. Index Terms--Analog model, digital model, impulse test, support vector machine (SVM), time-frequency analysis, transformer insulation, wavelet packet (WP) transform, wavelet transform (WT).

Abstract: The local Principal Component Analysis (PCA) reduces linearly redundant components that may present in higher dimensional space. It deploys an initial guess technique which can be utilized when the [...]

Abstract: Rijndael-The AES [1-3] is 128-bit block cipher based on an elegant algebraic structure over [F.sub.2.sup.8]. This cipher employs a simple approach to its substitution, permutation (SP) operations. We take [...]

The support vector machine (SVM) is a method for classification and for function approximation. This method commonly makes use of an [epsilon]-insensitive cost function, meaning that errors smaller than [epsilon] remain unpunished. As an alternative, a least squares support vector machine (LSSVM) uses a quadratic cost function. When the LSSVM method is used for function approximation, a nonsparse solution is obtained. The sparseness is imposed by pruning, i.e., recursively solving the approximation problem and subsequently omitting data that has a small error in the previous pass. However, omitting data with a small approximation error in the previous pass does not reliably predict what the error will be after the sample has been omitted. In this paper, a procedure is introduced that selects from a data set the training sample that will introduce the smallest approximation error when it will be omitted. It is shown that this pruning scheme outperforms the standard one. Index Terms--Function approximation, pruning, regression, support vector machine (SVM).

The electromagnetic field generated by a current dipole situated at an arbitrary position inside a conducting sphere is derived using the expansions of the spherical vector wave functions. The first few terms in a series expansion of this field with respect to the frequency are given for the normal magnetic field (used in magnetoencephalogram) and the tangential electric field (used in electroencephalogram) outside the conducting sphere at low frequency. It is shown that the first correction term to the static solution is linear in the frequency [Omega] (the second correction term is proportional to [[Omega].sup.3/2]) and, thus, the static solution can be used as a good approximation for the solution at a very low frequency. Index Terms - Conducting sphere, current dipole, electroencephalogram, expansion, frequency series, low-frequency, magnetoencephalogram, spherical vector wave functions.

A study was conducted to analyze an optimization problem for a general form of constrained-storage vector quantization (VQ), a coding method wherein the rate-distortion bound can be determined by increasing vector dimension. Signal compression applications associated with VQ were modeled using separate source codebook. The collection of vectors was also considered in the study using a codebook as an independent source while a data set of images was utilized for simulations.

A study was conducted to characterize a vector decoding method with a summation structure for reconstructing enhanced quality pictures from bitstreams produced by a standard transform encoder. An iterative algorithm was determined for designing locally optimal codebooks from a set of representative images. Results indicated that the vector decoder can be utilized with low complexity of computation and a codebook storage.

Triangular prism elements are useful in numerical solutions of electromagnetic field problems since they permit a three-dimensional (3-D) geometry to be generated by the extrusion of a triangular mesh. To date, however, few applications have employed vector basis functions on prism elements and the extension to distorted prisms reported in the literature apparently does not ensure cell-to-cell continuity. In this paper, we define interpolatory higher order curl- and divergence-conforming vector basis functions of the Nedelec type on prism elements, with extension to curved prisms, and discuss their completeness properties. Vector bases of arbitrary polynomial order are given and various results to confirm the faster convergence of higher order functions are presented. Index Terms - Electromagnetic fields, finite-element methods, numerical analysis.

The problem of scattering from two-dimensional (2-D) optical waveguide facets is analyzed. The facet is angled in vertical and horizontal planes, requiring a vectorial approach. This is achieved using the highly efficient Free Space Radiation Mode method, extended to the general vector case. The novel expressions for the reflectivity and coupling coefficients can be evaluated rapidly for several practical structures.

A new model is described that interprets the distribution of image data vectors using a modified version of Hunt's image model. An NNS algorithm that uses the results of the vector distribution model is presented and the relationship between the average of image blocks or vectors and the characteristics of image data is investigated. A possible explanation of image vector distribution is discussed.

A modification of the self-organizing feature maps algorithm improved its generalization abilities and produced a hidden Markov model (HMM) classifier preprocessor for determining machine tool dullness through vibration patterns. A clustering algorithm based on HMM was developed to group vector-series patterns and integrate information about desirable or probable class types. Testing with the vibration patterns produced by a drill demonstrated the viability of the approach.

Three steering vectors having distinct directions of arrival (DOA) are linearly independent for an electromagnetic vector sensor. Four steering vectors with distinct DOA's are linearly independent in case of distinct ellipticity angles between any two of the steering vectors. Five steering vectors are linearly independent when two or three of them correspond to circularly polarized signals with the same spin direction. Every four steering vectors corresponding to circularly polarized signals with same spin direction are linearly dependent.

The design of a class of learning vector quantizers (LVQ) involves the use multivariate order statistics that shows robust properties. The median LVQ is a special class of LVQ that uses either the marginal median or the vector median as a multivariate estimator of location. The performance of marginal median LVQ in color image quantization, in the presence of noise, is superior to that of classical LVQ. The execution time of the algorithm is also short.

The use of vector fields for color image processing is essential because of the correlation between the image channels. Vector directional filters (VDF) process image vectors by using directional information. VDF preserves the chromaticity component of images. A study considers issues related to the application of VDF and proposes efficient filtering schemes based on VDF, including adaptive and double-window structures.

An analysis of the signal-to-interference-plus-noise ratio (SINR) at the output of the minimum variance beamformer in the presence of steering vector errors provides an expression for the SNIR. The expressions for the weight vector and the general expression for the SNIR form the basis for the result presented. The special case of interference-related signal is studied. Steering vector errors, additive noise, the correlation between signal and interference, and finite-sample-size contribute to lower the performance. Simulations verify the analytical results.

A steady-state analysis of the multistage constant modulus (CM) array based on a Wiener formulation of the convergence condition shows that the multistage CM array can recover many cochannel sources and provide estimates of their angles of arrival. The closed-form expressions for the CM array and canceller weight vectors suggest that the unique canceller weights are proportional to the direction vectors of the captured sources. Computer simulations verify the results and demonstrate the impermanence of a cascade adaptive beamforming system for fixed and fading channels.

A wavelet-based image coding technique employing nonlinear interpolative vector quantization reduces the coding computational complexity. Appropriate coefficients of wavelet subimages were used to form 64-D vectors for three stages of decomposition. Vector quantization is performed on 16-D vectors extracted from the 64-D vectors. A nonlinear interpolative technique is used to reconstruct the 64-D vectors at the decoder. The method gives reconstructed images of acceptable quality even at compression ratios of 40:1 for gray images and 100:1 for color images.

The tree-structured nearest neighbor encoding method simplifies the full-search method without increasing distortion. The method uses algorithms to design a binary tree codebook, which is then used for nearest neighbor encoding. The short execution time permits slowly adaptive vector quantization techniques.

A coding scheme using the scalar-vector quantizer (SVQ) effectively compresses medical images. There is good rate-distortion performance of the SVQ, which is a fixed-rate encoder. The fixed-rate quantizer will probably reduce the complexities of variable-length scalar quantizers. The SVQ-based coder is an efficient method for picture archiving and communication systems, as their video monitor images look identical.

A derailment-free (DF), finite-state vector quantization (FSVQ) algorithm using state space optimization is an effective method for low bit rate image coding. The DF-FSVQ algorithm is better in performance, processing speed and system complexity than other FSVQ algorithms. The reduction in state space permits higher-order FSVQ.

A method that uses the advantages of subband coding, vector quantization, threshold vector quantization (TVQ) and finite-state vector quantization (FSVQ) improves the signal-to-noise ratio performance for encoding images. The technique uses correlation of subband signals. The encoding system uses TVQ in the higher subbands, with intraband and interband correlations. The LL(sub 1) band is encoded by the FSVQ method using high frequency bands.

An algorithm for creating finite sets of adaptive predictors using a cluster technique improves the 2-D-code excited linear prediction (CELP) coding of still images. The adaptive predictors are applied in a 2-D-CELP system with a codebook containing blocks of different sizes. The size of the block, number of predictors and order of predictors is balanced to the prediction gain. The method is better than the fixed block-size method.

The introduction of a tree-structured posterior estimator into the synergistic Bayes vector quantization (VQ) system improves classification with little change in the compression of two different image sources. The performance of many VQ-based algorithms used to get compression and classification of image sources is analyzed. The best method is a nonparametric method that puts a Bayes risk component in the distortion measure.

The encoding of still images using an adaptive edge-based side-match finite-state classified vector quantizer (FSVQ) is better than that by the side-match finite-state vector quantizer (SMVQ). The edge information is represented by a quadtree map in FSVQ. The improvement in image quality is up to 1.16 dB. FSVQ divides the blocks into edge and nonedge types. The edge blocks are again classified into subclasses depending on the edge-type of the neighbouring block.

A fractal image compression partitioning scheme based on Delaunay triangles gives less blocks than the square-based partition, which reduces the number of mappings. The partition is calculated on an initial set of points derived with a split and merge algorithm. The encoding complexity can also be reduced by a classification method based on vector quantization. The method is a modified version of the Lloyd algorithm.

An entropy-coded lattice-based vector quantizer (VQ) and noiseless code are suitable for image transform and subband image coding. The quantization is easy to use and the storage of vector codebooks is unnecessary. A Laplacian model is used as a basis for construction of the code. The code uses the classification of image block activity. The efficiency of coding is comparable to the best transform and coding methods being used.

The high-order conditional entropy-constrained vector quantization (CEC-RVQ) codes images better than the EC-RVQ. The CEC-RVQ algorithm has a better performance in rate-distortion with lesser coding complexity and memory requirements. The compression performance of CEC-RVQ is superior to that of EC-RVQ. The improvement in performance is due to the technique of finding the optimum conditioning stage symbols for a limited coder complexity and memory. The algorithm can be used with progressive transmission also.

The image quality obtained using a vector quantization technique for frame buffers with color palettes coded in the YIQ mode is better than that obtained with the RGB mode. The YIQ mode reduces the size of the color palette and the frame buffer use of a color image. The decrease in buffer size and color palette does not increase the complexity in the color calculations, or the logic or control of the frame buffer architecture.

The successive approximation wavelet vector quantization (SA-W-VQ) coding method allows the successive approximation of wavelet coefficients. The technique involves the coding of each vector by a decreasing vector series, to a specific distortion level. The SA-W-VQ is as efficient as some of the most successful coding systems.

The feature-specific vector quantizer uses a coding algorithm that compresses data beyond the second order decorrelation. The algorithm is based on studies of natural images and biological vision processing. The technique involves an initial hierarchical, size- and orientation-selective analytic bandpass decomposition. The decomposition is followed by transformation into amplitude and phase, and feature-specific partitioning of the derived orientation vector space.

A video coding algorithm based on a hybrid DPCM-vector quantization algorithm efficiently compresses test sequences at bit rates between 8 and 16 Kb/s. The algorithm provides good signal-to-noise ratios and subjective quality. The method involves the segmentation of difference images into blocks of variable sizes and shapes, and performs motion compensation and segmentation at the same time. Very active blocks require treatment by the fuzzy classified vector quantizer and large-size blocks are decimated to simplify their vector quantization.