Pattern Classification

This paper introduces the concept of distributed representation of fuzzy rules and applies it to classification problems. Distributed representation is implemented by superimposing many fuzzy rules corresponding to different fuzzy partitions of a pattern space. This means that we simultaneously employ many fuzzy rule tables corresponding to different fuzzy partitions in fuzzy inference. In order to apply distributed representation of fuzzy rules to pattern classification problems, we first propose an algorithm to generate fuzzy rules from numerical data. Next we propose a fuzzy inference method using the generated fuzzy rules. The classification power of distributed representation is compared with that of ordinary fuzzy rules which can be viewed as local representation.

In this paper, our goal is to (a) survey some of the legal contexts within which violence risk assessment already plays a prominent role, (b) explore whether developments in neuroscience could potentially be used to improve our ability to predict violence, and (c) discuss whether neuropredictive models of violence create any unique legal or moral problems above and beyond the well worn problems already associated with prediction more generally. In "Violence Risk Assessment and the Law", we briefly examine the role currently played by predictions of violence in three high stakes legal contexts: capital Neuroethics (2012) 5:67-99

A rigorous investigation on the synergy of mechanical attributes to engineer tactics for measuring human activity in terms of forces, as well as to provide independency and discrimination clarity of action recognition using linear and non-linear classification methodologies from data mining and evolutionary computation, are the main objectives where this paper focuses on. Mechanical analysis is employed to mathematically describe and model human movement by using a number of mechanical features inspired mainly from Kinematics Dynamics. Such features employ a twofold role on the descriptive analysis of an activity, initially to provide statistics regarding inertial expressions, probable hazard levels, body-status of energy loss, and finally to exploit these attributes by decomposing the 3D time series data for pattern recognition in terms of actions and behaviours. The performance statistics are being utilized by a mobile robot for remote surveillance within a smart environment.

natural adaptations or compensations. These should not be indicative of a deficient gait or be misconstrued as some age-related pathology. Objectives: To distinguish the gait patterns of young subjects from those of elderly men using three-dimensional (3D) gait data, to determine if elderly subjects displayed other than a typical gait pattern, and to identify which parameters best describe them.

In this paper we deal with the problem of feature selection by introducing a new approach based on Gravitational Search Algorithm (GSA). The proposed algorithm combines the optimization behavior of GSA together with the speed of Optimum-Path Forest (OPF) classifier in order to provide a fast and accurate framework for feature selection. Experiments on datasets obtained from a wide range of applications, such as vowel recognition, image classification and fraud detection in power distribution systems are conducted in order to asses the robustness of proposed technique against Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and a Particle Swarm Optimization (PSO)-based algorithm for feature selection.

The classification task for a real world application shall include a confidence estimation to handle unseen patterns i.e., patterns which were not considered during the learning stage of a classifier. This is important especially for safety critical applications where the goal is to assign these situations as "unknown" before they can lead to a false classification. Several methods were proposed in the past which were based on choosing a threshold on the estimated class membership probability. In this paper we extend the use of Gaussian mixture model (GMM)to estimate the uncertainty of the estimated class membership probability in terms of confidence interval around the estimated class membership probability. This uncertainty measure takes into account the number of training patterns available in the local neighborhood of a test pattern. Accordingly, the lower bound of the confidence interval or the number of training samples around a test pattern, can be used to detect the unseen patterns. Experimental results on a real-world application are discussed.

Finding sensitive and appropriate technologies for non-invasive observation and early detection of Alzheimer's disease (AD) is of fundamental importance to develop early treatments. In this work we develop a fully automatic computer aided diagnosis (CAD) system for high-dimensional pattern classification of baseline 18 F-FDG PET scans from Alzheimer's disease neuroimaging initiative (ADNI) participants. Image projection as feature space dimension reduction technique is combined with an eigenimage based decomposition for feature extraction, and support vector machine (SVM) is used to manage the classification task. A two folded objective is achieved by reaching relevant classification performance complemented with an image analysis support for final decision making. A 88.24% accuracy in identifying mild AD, with 88.64% specificity, and 87.70% sensitivity is obtained. This method also allows the identification of characteristic AD patterns in mild cognitive impairment (MCI) subjects. ]. Lower glucose metabolism has been correlated with dementia severity, either by assessing the correlation between performance on measures of overall cognitive function, such as the mini-mental state exam (MMSE), and cerebral metabolic rate for glucose (CMRgl) reduction over all study participants, or by comparing subgroups based on measures of disease severity, such as the MMSE, clinical dementia rating (CDR), or Global Deterioration Scale score .

Learning classifier systems (LCSs) are a machine learning technique, which combine reinforcement learning and evolutionary algorithms to evolve a set of classifiers (or rules) for pattern classification tasks. Despite promising performance across a variety of data sets, the performance of LCS is often degraded when data sets of high dimensionality and relatively few instances are encountered, a common occurrence with

The basic question of how to optimally make use of a finite number of available samples in designing pattern recognition systems is considered. This has several components: optimal use of the samples for design and testing; and the relationship between the number of measurements and the number of samples for various probability structural constraints. A spectrum of possibilities has been demonstrated, placing several apparently conflicting recent results in perspective.

An array of Love-wave sensors based on quartz and Novolac has been developed to detect chemical warfare agents (CWAs). These weapons are a risk for human health due to their efficiency and high lethality; therefore an early and clear detection is of enormous importance for the people safety. Love-wave devices realized on quartz as piezoelectric substrate and Novolac as guiding layer have been used to make up an array of six sensors, which have been coated with specific polymers by spin coating. The CWAs are very dangerous and for safety reasons their well known simulants have been used: dimethylmethyl phosphonate (DMMP), dipropyleneglycol methyl ether (DPGME), dimethylmethyl acetamide (DMA), dichloroethane (DCE), dichloromethane (DCM) and dichloropentane (DCP). The array has been exposed to these CWA simulants detecting very low concentrations, such as 25 ppb of DMMP, a simulant of nerve agent sarin. Finally, principal component analysis (PCA) as data pre-processing and discrimination technique, and probabilistic neural networks (PNN) as patterns classification technique have been applied. The performance of the sensor array has shown stability, accuracy, high sensitivity and good selectivity to these simulants.

A new non-linear Recursive Least Squares (RLS) algorithm is presented in the context of pattern classification problems. The algorithm incorporates the non-linearity of the filter's output in the updating rules of the classical RLS algorithm. The proposed method yields lower stationary error levels when compared to the standard LMS and RLS algorithms in a classical application of pattern classification, such as the channel equalization problem.

Fuzzy-Neural system has been applied to many engineering tasks. Fuzzy neurons in pattem classification are extremely useful because they provide a degree of membership information instead of numerical critic values such as "0" (bad) or "1 " (good). This paper shows a neural network application for automatic classijication of wooden boards. The basic processing unit consists of two types of generic OR and AND neurons structured in a four layer topology.

Behavioral information ac~luisition strategies in a diagnosis problem are formulated and examined and the effects on these strategies of both information presentation and user training are tested. Twenty-five subjects participated in an experiment in which the decision maker was required to solve a set of medical diagnosis problems. A matrix of 7 illnesses and 11 po~ble tests was presented on a micro-computer; the task was to diagnose the correct illness with the minimum number of tests. The results were analyzed using verbal protocols and a computerized record of the user's responses. The results confirm the use of certain behavioral strategies, in particular concentration on a small subset of classes in a regular pattern, and a positional effect on the choice of tests was also detected. The implications of these results on the design of management information ~ems are discussed.

A direct, featureless process to classify contact impressions of objects gripped by a robot hand is presented. The inforlnation about the type of contact allows the selection of the most appropriate manipulating strategy to handle the grasped object. A learning vector quantization (LVQ) network is applied introducing a contact-pattern preprocessing technique to improve the robustness of tbe classification with respect to the pattern variations in position, orientation and size. © 1997 Elsevier Science S.A.

The security of computer networks plays a strategic role in modern computer systems. In order to enforce high protection levels against threats, a number of software tools have been currently developed. Intrusion Detection Systems aim at detecting intruders who elude ''first line'' protection. In this paper, a pattern recognition approach to network intrusion detection based on the fusion of multiple classifiers is proposed. Five decision fusion methods are assessed by experiments and their performances compared. The potentialities of classifier fusion for the development of effective intrusion detection systems are evaluated and discussed.

The k-nearest neighbor rule is one of the simplest and most attractive pattern classification algorithms. However, it faces serious challenges when patterns of different classes overlap in some regions in the feature space. In the past, many researchers developed various adaptive or discriminant metrics to improve its performance. In this paper, we demonstrate that an extremely simple adaptive distance measure significantly improves the performance of the k-nearest neighbor rule.

Doppler umbilical artery blood flow velocity waveform measurements are used in perinatal surveillance for the evaluation of fetal condition. There is an ongoing debate on the predictive value of Doppler measurements concerning the critical effect of the selection of parameters for the interpretation of Doppler waveforms. In this paper, we describe how neural network methods can be used both to discover relevant classification features and subsequently to classify Doppler umbilical artery blood flow velocity waveforms. Results obtained from 199 normal and high risk patients' umbilical artery waveforms highlighted a classification concordance varying from 90 to 98% accuracy.

Hierarchical Incremental Class Learning (HICL) is a new task decomposition method that addresses the pattern classification problem. The HICL is proven to be a good classifier but closer examination reveals areas for potential improvement. This paper proposes a theoretical model to evaluate the performance of HICL and presents an approach to improve the classification accuracy of HICL by applying the concept of Reduced Pattern Training (RPT). The theoretical analysis shows that HICL can achieve better classification accuracy than Output Parallelism [Guan and Li: IEEE Transaction on Neural Networks, 13 (2002), 542–550]. The procedure for RPT is described and compared with the original training procedure. The RPT reduces systematically the size of the training data set based on the order of sub-networks built. The results from four benchmark classification problems show much promise for the improved model.

Recognizing people by gait promises to be useful for identifying individuals from a distance; in this regard, improved techniques are under development. In this paper, an improved method for gait recognition is proposed. Binarized silhouette of a motion object is first represented by four 1-D signals that are the basic image features called the distance vectors. The distance vectors are differences between the bounding box and silhouette, and extracted using four projections to silhouette. Fourier Transform is employed as a preprocessing step to achieve translation invariant for the gait patterns accumulated from silhouette sequences that are extracted from the subjects’ walk in different speed and/or different time. Then, eigenspace transformation is applied to reduce the dimensionality of the input feature space. Support vector machine (SVM)-based pattern classification technique is then performed in the lower-dimensional eigenspace for recognition. The input feature space is alternatively constructed by using two different approaches. The four projections (1-D signals) are independently classified in the first approach. A fusion task is then applied to produce the final decision. In the second approach, the four projections are concatenated to have one vector and then pattern classification with one vector is performed in the lower-dimensional eigenspace for recognition. The experiments are carried out on the most well-known public gait databases: the CMU, the USF, SOTON, and NLPR human gait databases. To effectively understand the performance of the algorithm, the experiments are executed and presented as increasing amounts of the gait cycles of each person available during the training procedure. Finally, the performance of the proposed algorithm is comparatively illustrated to take into consideration the published gait recognition approaches.

For the last two decades, lot of research has been done on neural networks, resulting in many types of neural networks. These neural networks can be implemented in number of ways. Due to the revival of research interest in neural networks, some important technological developments have been made in VLSI. This paper discusses comparative study between analog implementation and digital implementation for neural networks. The discussion topics include power-consumption, area, robustness, and implementation efficiency of these implementation techniques respectively. It can be estimated whether an analog or digital neural network is optimum for a specific application. It is observed that the choice between analog and digital neural networks is application dependent. The goal is to estimate which type of implementation should be used for which class of applications. This work is based on the study of neural implementations restricted only to pattern classification and focuses on widely used layered feed-forward neural network.

Certain diseases cause permanent changes to the shapes and densities of nailfold capillaries and, therefore, nailfold capillaroscopy is important as a tool for diagnosing and monitoring these diseases. The first aim of the project is to resolve differences in terminology that have developed over the years in previous work. We propose a taxonomy for nailfold capillaries that cover six descriptive classes: cuticulis, open, tortuous, crossed, bushy, and bizarre. The first three are parametric in that they may be distinguished by the ratio of capillary length to width and by the curvature of the capillary limbs. The last three are characterized by their topology; a crossed capillary has a closed area that is not connected to the image background. Bushy and bizarre capillaries have atypical shapes that are characterized by the convex hull of their skeleton. These descriptive classes may be modified according to anomalies in width and length.

The Liquid State Machine (LSM) is a recently developed computational model with interesting properties. It can be used for pattern classification, function approximation and other complex tasks. Contrary to most common computational models, the LSM does not require information to be stored in some stable state of the system: the inherent dynamics of the system are used by a memoryless

Artificial neural network based technology, which is inspired by biological neural networks, has developed rapidly in the previous decade and has been applied in power system protection applications. Protection of transmission and sub transmission lines is most frequently done using distance relays. This paper discusses the design, implementation and testing of an artificial neural network (ANN) based distance relay that implements characteristics of a mho (admittance) relay. The Artificial Neural Network based designs of distance relays proposed so far work well for faults well within the boundary characteristics of a relay, but are not able to maintain the integrity of the boundaries of the relay characteristics of generic designs.

The Time Adaptive SOM, or TASOM, is used to automatically adjust learning rate and neighborhood size of each neuron of the SOM network independently. Each neuron's learning rate is determined by a function of the distance between an input vector and its weight vector. The width ofthe neighborhood function is updated by a function of the distance between the weight vector of the neuron and the weight vectors of neighboring neurons. Only one time parameter initialization is sufficient throughout the lifetime of TASOM to work in stationary and nonstationary environments without retraining. In this paper, the TASOM is tested with standard data sets including the iris plant, breast cancer, and BUPA liver disease data for classification of input vectors. The tests carried out in stationary and non-stationary environments demonstrate that the TASOM can work for classification without the need for reinitializing the network parameters and weights.

Digital image processing has the potential to support the identification of plant species required for site-specific weed control in grassland swards. The present study focuses on the identification of one of the most invasive and persistent weed species on European grassland, the broad-leaved dock (Rumex obtusifolius L., R.o.), in complex mixtures of perennial ryegrass with R.o. and other herbs.

We develop a face recognition algorithm which is insensitive to large variation in lighting direction and facial expression. Taking a pattern classification approach, we consider each pixel in an image as a coordinate in a high-dimensional space. We take advantage of the observation that the images of a particular face, under varying illumination but fixed pose, lie in a 3D linear subspace of the high dimensional image space-if the face is a Lambertian surface without shadowing. However, since faces are not truly Lambertian surfaces and do indeed produce self-shadowing, images will deviate from this linear subspace. Rather than explicitly modeling this deviation, we linearly project the image into a subspace in a manner which discounts those regions of the face with large deviation. Our projection method is based on Fisher's Linear Discriminant and produces well separated classes in a low-dimensional subspace, even under severe variation in lighting and facial expressions. The Eigenface technique, another method based on linearly projecting the image space to a low dimensional subspace, has similar computational requirements. Yet, extensive experimental results demonstrate that the proposed "Fisherface" method has error rates that are lower than those of the Eigenface technique for tests on the Harvard and Yale Face Databases.

A multi-scale, morphological method for the purpose of shape-based object recognition is presented. A connected operator similar to the morphological hat-transform is defined, and two scale-space representations are built, using the curvature function as the underlying 1-D signal. Each peak and valley of the curvature is extracted and described by its maximum and average heights and by its extent, and represents an entry in the top or bottom hat-transform scale spaces. We demonstrate object recognition based on hat-transform scale spaces for three large data sets, a set of diatom contours, the set of silhouettes from the MPEG-7 database and the set of 2-D views of 3-D objects from the COIL-20 database. Our approach outperforms other methods for which comparative results exist.

We present an improved unbiased algorithm for determining principal curves in high dimensional spaces, and then propose two novel applications of principal curve to feature extraction and pattern classification–the Principal Curve Feature Extractor (PCFE) and the Principal ...

We observed that the two costly optimization problems of twin support vector machine (TWSVM) classifier can be avoided by introducing a technique as used in proximal support vector machine (PSVM) classifier. With this modus operandi we formulate a much simpler nonparallel plane proximal classifier (NPPC) for speeding up the training of it by reducing significant computational burden over TWSVM. The formulation of NPPC for binary data classification is based on two identical mean square error (MSE) optimization problems which lead to solving two small systems of linear equations in input space. Thus it eliminates the need of any specialized software for solving the quadratic programming problems (QPPs). The formulation is also extended for nonlinear kernel classifier. Our computations show that a MATLAB implementation of NPPC can be trained with a data set of 3 million points with 10 attributes in less than 3 s. Computational results on synthetic as well as on several bench mark data sets indicate the advantages of the proposed classifier in both computational time and test accuracy. The experimental results also indicate that performances of classifiers obtained by MSE approach are sufficient in many cases than the classifiers obtained by standard SVM approach.

Excitation-continuous music instrument control patterns are often not explicitly represented in current sound synthesis techniques when applied to automatic performance. Both physical model-based and sample-based synthesis paradigms would benefit from a flexible and accurate instrument control model, enabling the improvement of naturalness and realism. We present a framework for modeling bowing control parameters in violin performance. Nearly non-intrusive sensing techniques allow for accurate acquisition of relevant timbre-related bowing control parameter signals. We model the temporal contour of bow velocity, bow pressing force, and bow-bridge distance as sequences of short Bézier cubic curve segments. Considering different articulations, dynamics, and performance contexts, a number of note classes are defined. Contours of bowing parameters in a performance database are analyzed at note-level by following a predefined grammar that dictates characteristics of curve segment sequences for each of the classes in consideration. As a result, contour analysis of bowing parameters of each note yields an optimal representation vector that is sufficient for reconstructing original contours with significant fidelity. From the resulting representation vectors, we construct a statistical model based on Gaussian mixtures suitable for both the analysis and synthesis of bowing parameter contours. By using the estimated models, synthetic contours can be generated through a bow planning algorithm able to reproduce possible constraints caused by the finite length of the bow. Rendered contours are successfully used in two preliminary synthesis frameworks: digital waveguide-based bowed string physical modeling and sample-based spectral-domain synthesis.

Uncertainty due to spatial variability of hydraulic conductivity is an important issue in the design of reliable groundwater remediation strategies. Using groundwater management models based on a stochastic approach to groundwater flow, where the log-hydraulic conductivity is represented as a random field, is a frequently studied technique for the design of aquifer remediation in the presence of uncertainty. Such an approach employs the solution of a management model for a large set of equally probable realizations of the hydraulic conductivity. However, only a few out of the large set of realizations are critical to the final outcome of the design. The spatial distribution of the hydraulic conductivity values in a realization, and the degree of variation of the hydraulic conductivity values within a realization are identified as two important features that determine the level of criticalness of a realization. The association between the hydraulic conductivity pattern and the level of criticalness is not known explicitly and needs to be captured for efficient screening. The screening approach presented here utilizes the pattern classification capability of a neural network and its ability to learn from examples. It is shown that incorporation of only a few critical realizations in a groundwater management model can yield highly reliable remediation designs. The application of the screening tool in a pump-and-treat design problem is illustrated via two examples. INTRODUCTION The effectiveness of groundwater monitoring and reclamation designs depends on the availability of knowledge concerning the parameters that influence groundwater flow and subsurface contaminant transport. Accurate prediction of the actual patterns of groundwater flow and contaminant transport is an impossible task given the natural variability of the subsurface properties. One of the major sources of uncertainty is the spatial variability of the hydraulic conductivity of the porous media. Although acquiring complete knowledge of hydrogeologic parameters is extremely costly, less costly partial information can provide estimates of the statistical properties of the porous media. This limited knowledge enables the conceptualization of the uncertain hydraulic conductivity using equally likely realizations of a random field with known statistical properties. Many researchers have adopted stochastic approaches in groundwater management models or monitoring network designs [

In this study, we developed an automated scheme to classify the synoptic meteorological conditions governing over the northern Thailand during the winter period (November–January) into six-synoptic patterns representing different development stages of the regional pressure ...

A new algorithm for massive lesion detection in mammography is presented. The algorithm consists in three main steps : 1) reduction of the dimension of the image to be processed through the identifi cation of regions of interest (rois) as candidates for massive lesions ; 2) characterization of the roi by means of suitable feature extraction ; 3) pattern classifi cation through supervised neural networks.

Automatic classification of lung tissue patterns in high resolution computed tomography images of patients with interstitial lung diseases is an important stage in the construction of a computer-aided diagnosis system. To this end, a novel approach is proposed using two sets of overcomplete wavelet filters, namely discrete wavelet frames (DWF) and rotated wavelet frames (RWF), to extract the features which best characterizes the lung tissue patterns. Support vector machines learning algorithm is then applied to perform the pattern classification. Four different lung patterns (ground glass, honey combing, reticular, and normal) selected from a database of 340 images are classified using the proposed method. The overall multiclass accuracy reaches 90. 72%, 95.85%, and 96.81% for DWF, RWF, and their combination, respectively. These results prove that RWF is superior to DWF, due to its orientation selectivity. However, best results are obtained by the combination of two filter banks which shows that the two set of filters are complementary.

One of the important problems in medical imaging is twoclass classification, for example determination of benign from malignant cases in breast cancer treatment. In this paper we present a new support vector machine method for two-class medical image classification. The key idea of this method is to construct an optimal hypersphere such that both the interior margin between the surface of this sphere and the normal data, and the exterior margin between this surface and the abnormal data are as large as possible. The proposed method is easily implemented and can reduce both false positive and false negative error rates to obtain very good classification results. Experiments were performed on three medical image data sets to evaluate the proposed method.

We have demonstrated powerful new techniques for identifying the optical impairments causing the degradation of an optical channel. We use machine learning and pattern classification techniques on eye diagrams to identify the optical impairments. These capabilities can enable the development of low-cost optical performance monitors having significant diagnostic capabilities.

Feature selection is an important step in many pattern classification problems. It is applied to select a subset of features, from a much larger set, such that the selected subset is sufficient to perform the classification task. Due to its importance, the problem of feature selection has been investigated by many researchers. In this paper, a novel feature subset search procedure that utilizes the Ant Colony Optimization (ACO) is presented. The ACO is a metaheuristic inspired by the behavior of real ants in their search for the shortest paths to food sources. It looks for optimal solutions by considering both local heuristics and previous knowledge. When applied to two different classification problems, the proposed algorithm achieved very promising results.

The majority of multi-class pattern classification techniques are proposed for learning from balanced datasets. However, in several real-world domains, the datasets have imbalanced data distribution, where some classes of data may have few training examples compared for other classes. In this paper we present our research in learning from imbalanced multi-class data and propose a new approach, named Multi-IM, to deal with this problem. Multi-IM derives its fundamentals from the probabilistic relational technique (PRMs-IM), designed for learning from imbalanced relational data for the two-class problem. Multi-IM extends PRMs-IM to a generalized framework for multi-class imbalanced learning for both relational and non-relational domains.

The problem of Automatic Fingerprint Pattern Classification (AFPC) has been studied by many fingerprint biometric practitioners. It is an important concept because, in instances where a relatively large database is being queried for the purposes of fingerprint matching, it serves to reduce the duration of the query. The fingerprint classes discussed in this document are the Central Twins (CT), Tented Arch (TA), Left Loop (LL), Right Loop (RL) and the Plain Arch (PA). The classification rules employed in this problem involve the use of the coordinate geometry of the detected singular points. Using a confusion matrix to evaluate the performance of the fingerprint classifier, a classification accuracy of 83.5% is obtained on the five-class problem. This performance evaluation is done by making use of fingerprint images from one of the databases of the year 2002 version of the Fingerprint Verification Competition (FVC2002).

Here we describe a functional magnetic resonance imaging study of humans engaged in memory search during a free recall task. Patterns of cortical activity associated with the study of three categories of pictures (faces, locations, and objects) were identified by a pattern-classification algorithm. The algorithm was used to track the reappearance of these activity patterns during the recall period. The reappearance of a given category's activity pattern correlates with verbal recalls made from that category and precedes the recall event by several seconds. This result is consistent with the hypothesis that category-specific activity is cueing the memory system to retrieve studied items.

Support vector machine (SVM) is a novel pattern classification method that is valuable in many applications. Kernel parameter setting in the SVM training process, along with the feature selection, significantly affects classification accuracy. The objective of this study is to obtain the better parameter values while also finding a subset of features that does not degrade the SVM classification accuracy. This study develops a simulated annealing (SA) approach for parameter determination and feature selection in the SVM, termed SA-SVM.

Subspace selection approaches are powerful tools in pattern classification and data visualization. One of the most important subspace approaches is the linear dimensionality reduction step in the Fisher's linear discriminant analysis (FLDA), which has been successfully employed in many fields such as biometrics, bioinformatics, and multimedia information management. However, the linear dimensionality reduction step in FLDA has a critical drawback: for a classification task with c classes, if the dimension of the projected subspace is strictly lower than c - 1, the projection to a subspace tends to merge those classes, which are close together in the original feature space. If separate classes are sampled from Gaussian distributions, all with identical covariance matrices, then the linear dimensionality reduction step in FLDA maximizes the mean value of the Kullback-Leibler (KL) divergences between different classes. Based on this viewpoint, the geometric mean for subspace selection is studied in this paper. Three criteria are analyzed: 1) maximization of the geometric mean of the KL divergences, 2) maximization of the geometric mean of the normalized KL divergences, and 3) the combination of 1 and 2. Preliminary experimental results based on synthetic data, UCI Machine Learning Repository, and handwriting digits show that the third criterion is a potential discriminative subspace selection method, which significantly reduces the class separation problem in comparing with the linear dimensionality reduction step in FLDA and its several representative extensions.

Real time Fault Detection and Diagnosis (FDD) is an important area of research interest in Knowledge Based Expert Systems. Neurocomputing is one of fastest growing areas of research in the fields of Artificial Intelligence and Pattern Recognition. This paper explores the suitability of pattern classification approach of neural networks for fault detection and diagnosis. Suitability of using neural network as pattern classifiers for power system fault diagnosis is described in detail. A neural network design and simulation environment for real-time FDD is presented. An Analysis of the learning, recall and generalization charecterstisc of the neural network diagnostic system is presented and discussed in detail.

The mixture of gait deviations seen in patients following a stroke is remarkably variable. An objective system for classification of gait patterns for this population could be used to guide treatment planning. Quantitated gait analysis was conducted for 47 individuals at admission to in-patient rehabilitation and again at 6 months post-stroke for 42 subjects. Non-hierarchical cluster analysis was used to classify the gait patterns of patients based on the temporal Á/spatial and kinematic parameters of walking. Four clusters of patients were identified at both assessment intervals. At the admission test walking velocity, peak knee extension in mid stance and peak dorsiflexion in swing were the three factors that best characterized the groups. At 6 months the explanatory variables were velocity, knee extension in terminal stance, and knee flexion in pre swing. Differences in muscle strength and muscle activation patterns during walking were identified between groups. #

Document Image processing and Optical Character Recognition (OCR) have been a frontline research area in the field of human-machine interface for the last few decades. Recognition of Indian language characters has been a topic of interest for quite some time. The earlier contributions were reported in [1] and [2]. A more recent work is reported in [3] and [9]. The need for efficient and robust algorithms and systems for recognition is being felt in India, especially in the post and telegraph department where OCR can assist the staff in sorting mail. Character recognition can also form a part in applications like intelligent scanning machines, text to speech converters, and automatic language-to-language translators.