Search

This paper presents two new methods of elastic articulated objects modeling and motion estimation based on a new revolving conic surface. The deformation of human body surfaces can be represented by adjusting only one or two deformation parameters for each limb. Then, the 3D deformation and motion parameters are estimated by corresponding 2D image points and contours from a sequences of stereo images. [ABSTRACT FROM AUTHOR]

This paper presents a method to recover the 3D configuration of a face in each frame of a video. The 3D configuration consists of the 3 translational parameters and the 3 orientation parameters which correspond to the yaw, pitch and roll of the face, which is important for applications like face modeling, recognition, expression analysis, etc. The approach combines the structural advantages of geometric modeling with the statistical advantages of a particle-filter based inference. The face is modeled as the curved surface of a cylinder which is free to translate and rotate arbitrarily. The geometric modeling takes care of pose and self-occlusion while the statistical modeling handles moderate occlusion and illumination variations. Experimental results on multiple datasets are provided to show the efficacy of the approach. The insensitivity of our approach to calibration parameters (focal length) is also shown. [ABSTRACT FROM AUTHOR]

Hand gesture recognition from visual images finds applications in areas like human computer interaction, machine vision, virtual reality and so on. Vision-based hand gesture recognition involves visual analysis of hand shape, position and/or movement. In this paper, we present a model-based method for tracking hand motion in a complex scene, thereby estimating the hand motion trajectory. In our proposed technique, we first segment the frames into video object planes (VOPs) with the hand as the video object. This is followed by hand tracking using Hausdorff tracker. In the next step, the centroids of all VOPs are calculated using moments as well as motion information. Finally, the hand trajectory is estimated by joining the VOP centroids. In our experiment, the proposed trajectory estimation algorithm gives about 99% accuracy in finding the actual trajectory. [ABSTRACT FROM AUTHOR]

In this paper, a method for improvement of image quality using the error diffusion which considers the APL process is proposed. In the proposed method, the APL process is performed before the error diffusion process. Simulation results showed that the proposed method has better performances for resolution in images and CCT uniformity according to each grayscale than the conventional method. [ABSTRACT FROM AUTHOR]

Differential Evolution (DE) is a simple and surprisingly efficient algorithm for global optimization over continuous spaces. It has reportedly outperformed many versions of EA and other search heuristics when tested over both benchmark and real world problems. However the performance of DE deteriorates severely if the fitness function is noisy and continuously changing. In this paper we propose an improved DE scheme which can efficiently track the global optima of a noisy function. The scheme performs better than the classical DE, PSO, and an EA over a set of benchmark noisy problems. [ABSTRACT FROM AUTHOR]

In this paper we propose a new Hybrid Particle Swarm Optimizer model based on particle swarm, with breeding concepts from novel evolutionary algorithms. The hybrid PSO combines traditional velocity and position update rules of RANDIW-PSO and ideas from Self Adaptive Pareto Differential Evolution Algorithm (SPDE). The hybrid model is tested and compared with some high quality PSO models like the RANDIW-PSO and TVIW-PSO. The results indicate two good prospects of our proposed hybrid PSO model: potential to achieve faster convergence as well as potential to find a better solution. The hybrid PSO model, with the abovementioned features, is then efficiently utilized to coordinate robot ants in order to help them to probe as much camera coverage area of some planetary surface or working field as possible with minimum common area coverage. [ABSTRACT FROM AUTHOR]

Determination of an isothetic polygonal approximation of the outer or inner contour of an object is a challenging problem with numerous applications to pattern recognition and image processing. In this paper, a novel algorithm is presented for constructing the outer (or inner) tight isothetic polygon(s) containing (or filling) an arbitrarily shaped 2D object on a background grid, using a classical labeling technique. The background grid may consist of uniformly or non-uniformly spaced horizontal and vertical lines. Experimental results for both uniform and non-uniform grids of varying sizes have been reported to demonstrate the applicability and efficiency of the proposed algorithm. [ABSTRACT FROM AUTHOR]

A holistic classification system for off-line recognition of legal amounts in checks is described in this paper. The binary images obtained from the cursive words are processed following the human visual system, employing a Hough transform method to extract perceptual features. Images are finally coded into a bidimensional feature map representation. Multilayer perpeptrons are used to classify these feature maps into one of the 32 classes belonging to the CENPARMI database. To select a final classification system, ROC graphs are used to fix the best threshold values of the classifiers to obtain the best tradeoff between accuracy and misclassification. [ABSTRACT FROM AUTHOR]

In this paper we present and analyze an artificial neural network hardware engine, its architecture and implementation. The engine was designed to solve performance problems of the serial software implementations. It is based on a hierarchical parallel and parameterized architecture. Taking into account verification results, we conclude that this engine improves the computational performance, producing speedups from 52.3 to 204.5 and its architectural parameterization provides more flexibility. [ABSTRACT FROM AUTHOR]

Performance of hierarchical classifiers depends on two aspects - the performance of the individual classifiers, and the design of the architecture. In this paper, we present a scheme for designing hybrid hierarchical classifiers under user specified constraints on time and space. [ABSTRACT FROM AUTHOR]

This paper presents an investigation on pattern classification techniques applied to voltage sag monitoring data. Similar pattern groups or sets of classes, resulting from a voltage sag classification, represent disturbance categories that may be used as indexes for a cause/effect disturbance analysis. Various classification algorithms are compared in order to establish a classifier design. Results over clustering performance indexes are presented for hierarchical, fuzzy c-means and k-means unsupervised clustering techniques, and a principal component analysis is used for features (or attributes) choice. The efficiency of the algorithms was analyzed by applying the CDI and DBI indexes. [ABSTRACT FROM AUTHOR]

As computer security becomes important, various system security mechanisms have been developed. Especially anomaly detection using hidden Markov model has been actively exploited. However, it can only detect abnormal behaviors under predefined threshold, and it cannot identify the type of intrusions. This paper aims to identify the type of intrusions by analyzing the state sequences using Viterbi algorithm and calculating the distance between the standard state sequence of each intrusion type and the current state sequence. Because the state sequences are not always extracted consistently due to environmental factors, edit distance is utilized to measure the distance effectively. Experimental results with buffer overflow attacks show that it identifies the type of intrusions well with inconsistent state sequences. [ABSTRACT FROM AUTHOR]

In this paper we propose a multistage unsupervised classifier which uses graph-cut to produce initial segments which are made up of pixels with similar spectral properties, subsequently labelled by a fuzzy c-means clustering algorithm into a known number of classes. These initial segmentation results are used as a seed to the expectation maximization (EM) algorithm. Final classification map is produced by using the maximum likelihood (ML) classifier, performance of which is quite good as compared to other unsupervised classification techniques. [ABSTRACT FROM AUTHOR]

This paper provides an overview of a number of techniques developed within our group to perform 3D reconstruction and image segmentation based of the application of energy minimisation concepts. We begin with classical snake techniques and show how similar energy minimisation concepts can be extended to derive globally optimal segmentation methods. Then we discuss more recent work based on geodesic active contours that can lead to globally optimal segmentations and reconstructions in 2D. Finally we extend the work to 3D by introducing continuous flow globally minimal surfaces. Several applications are discussed to show the wide applicability and suitability of these techniques to several difficult image analysis problems. [ABSTRACT FROM AUTHOR]

A new pattern recognition method for classification of multi dimensional samples is proposed. In pattern recognition problems samples (pixels in remote sensing) are described using a number of features (dimensions/bands in remote sensing). While a number of features of the samples are useful for a better description of the image, they pose a threat in terms of unwieldy mass of data. In this paper we propose a method to achieve dimensionality reduction using regression. The method proposed transforms the feature values into representative patterns, termed as symbolic objects, which are obtained through regression lines. The so defined symbolic object accomplishes dimensionality reduction of the data. A new distance measure is devised to measure the distances between the symbolic objects (fitted regression lines) and clustering is preformed. The efficacy of the method is corroborated experimentally. Keywords: Pattern Classification, Dimensionality Reduction, Feature Sequence, Regression, Clustering, Data Assimilation, Multi Dimensional Data, Symbolic Data. [ABSTRACT FROM AUTHOR]

In the typical nonparametric approach to classification in instance-based learning and data mining, random data (the training set of patterns) are collected and used to design a decision rule (classifier). One of the most well known such rules is the k-nearest neighbor decision rule (also known as lazy learning) in which an unknown pattern is classified into the majority class among the k-nearest neighbors in the training set. This rule gives low error rates when the training set is large. However, in practice it is desired to store as little of the training data as possible, without sacrificing the performance. It is well known that thinning (condensing) the training set with the Gabriel proximity graph is a viable partial solution to the problem. However, this brings up the problem of efficiently computing the Gabriel graph of large training data sets in high dimensional spaces. In this paper we report on a new approach to the instance-based learning problem. The new approach combines five tools: first, editing the data using Wilson-Gabriel-editing to smooth the decision boundary, second, applying Gabriel-thinning to the edited set, third, filtering this output with the ICF algorithm of Brighton and Mellish, fourth, using the Gabriel-neighbor decision rule to classify new incoming queries, and fifth, using a new data structure that allows the efficient computation of approximate Gabriel graphs in high dimensional spaces. Extensive experiments suggest that our approach is the best on the market. [ABSTRACT FROM AUTHOR]

In many application areas of imaging sciences, object information captured in multi-dimensional images needs to be extracted, visualized, manipulated, and analyzed. These four groups of operations have been (and are being) intensively investigated, developed, and applied in a variety of applications. In this paper, we put forth two main arguments: (1) Computers are digital, and most image acquisition and communication efforts at present are toward digital approaches. In the same vein, there are considerable advantages to taking an inherently digital approach to the above four groups of operations rather than using concepts based on continuous approximations. (2) Considering the fact that images are inherently fuzzy, to handle uncertainties and heterogeneity of object properties realistically, approaches based on fuzzy sets should be taken to the above four groups of operations. We give two examples in support of these arguments. [ABSTRACT FROM AUTHOR]

The Semantic Web is not a single technology, but rather a collection of technologies designed to work together. As a result, research on the Semantic Web intends both to advance individual technologies as well as to integrate them and take advantage of the result. In this paper we present new work on many layers of the Semantic Web, including content generation, web services, e-connections, and trust. [ABSTRACT FROM AUTHOR]

In this paper, we present an algorithm for detection and tracking of small objects, like a ping pong ball or a cricket ball in sports video sequences. It can also detect and track airborne targets in an infrared image sequence. The proposed method uses motion as the primary cue for detection. The detected object is tracked using the multiple filter bank approach. Our method is capable of detecting objects of low contrast and negligible texture content. Moreover, the algorithm also detects point targets. The algorithm has been evaluated using large number of different video clips and the performance is analysed. [ABSTRACT FROM AUTHOR]

Cluster analysis deals with the automatic discovery of the grouping of a set of patterns. Despite more than 40 years of research, there are still many challenges in data clustering from both theoretical and practical viewpoints. In this paper, we describe several recent advances in data clustering: clustering ensemble, feature selection, and clustering with constraints. [ABSTRACT FROM AUTHOR]

This paper addresses partially supervised classification problems, i.e. problems in which different data sets referring to the same scenario (phenomenon) should be classified but a training information is available only for some of them. In particular, we propose a novel approach to the partially supervised classification which is based on a Bi-transductive Support Vector Machines (T2-SVM). Inspired by recently proposed Transductive SVM (TSVM) and Progressive Transductive SVM (PTSVM) algorithms, the T2-SVM algorithm extracts information from unlabeled samples exploiting the transductive inference, thus obtaining high classification accuracies. After defining the formulation of the proposed T2-SVM technique, we also present a novel accuracy assessment strategy for the validation of the classification performances. The experimental results carried out on a real remote sensing partially supervised problem confirmed the reliability and the effectiveness of both the T2-SVM and the corresponding validation procedure. [ABSTRACT FROM AUTHOR]

Images constitute data that live in a very high dimensional space, typically of the order of hundred thousand dimensions. Drawing inferences from correlated data of such high dimensions often becomes intractable. Therefore traditionally several of these problems like face recognition, object recognition, scene understanding etc. have been approached using techniques in pattern recognition. Such methods in conjunction with methods for dimensionality reduction have been highly popular and successful in tackling several image processing tasks. Of late, the advent of cheap, high quality video cameras has generated new interests in extending still image-based recognition methodologies to video sequences. The added temporal dimension in these videos makes problems like face and gait-based human recognition, event detection, activity recognition addressable. Our research has focussed on solving several of these problems through a pattern recognition approach. Of course, in video streams patterns refer to both patterns in the spatial structure of image intensities around interest points and temporal patterns that arise either due to camera motion or object motion. In this paper, we discuss the applications of pattern recognition in video to problems like face and gait-based human recognition, behavior classification, activity recognition and activity based person identification. [ABSTRACT FROM AUTHOR]

The problem of finding association rules from a dataset is to find all possible associations that hold among the items, given a minimum support and confidence. This involves finding frequent sets first and then the association rules that hold within the items in the frequent sets. In temporal datasets as the time in which a transaction takes place is important we may find sets of items that are frequent in certain time intervals but not frequent throughout the dataset. These frequent sets may give rise to interesting rules but these can not be discovered if we calculate the supports of the item sets in the usual way. We call here these frequent sets locally frequent. Normally these locally frequent sets are periodic in nature. We propose modification to the Apriori algorithm to compute locally frequent sets and periodic frequent sets and periodic association rules. [ABSTRACT FROM AUTHOR]

Text mining techniques have been proposed for extracting protein names and their interactions. First, we have made improvements on existing methods for handling single word protein names consisting of characters, special symbols, and numbers. Second, compound word protein names are extracted using conditional probabilities of the occurrences of neighboring words. Third, interactions are extracted based on Bayes theorem over discriminating verbs that represent the interactions of proteins. Experimental results demonstrate the feasibility of our approach with improved performance. [ABSTRACT FROM AUTHOR]

In this paper we present a parallel algorithm for local alignment of two biological sequences. Given two sequences of size m and n, our algorithm uses a novel technique namely, carry lookahead and requires O(m / 4 + n / 2) time on a maximum of O(m) processors. [ABSTRACT FROM AUTHOR]

This paper compares the most utilized spread spectrum technique with the newly evolved technique based on Singular Value Decomposition (SVD) for watermarking digital images. Both techniques are tested for a variety of attacks and the simulation results show that the watermarks generated by these techniques have complimentary robustness properties. A new hybrid technique, combining both paradigms, is proposed that is capable of surviving an extremely wide range of attacks. An image is first watermarked using spread spectrum and then a SVD based watermark is added to the watermarked image. The resulting double watermarked image is extremely robust to a wide range of distortions. [ABSTRACT FROM AUTHOR]

A new split-approach based method for approximating a boundary curve by a polygon is proposed in this paper. The proposed method recursively splits boundary curve into smaller segment with the help of small eigenvalue of covariance matrix of the boundary curve. Set of boundary points at which boundary curve is split into smaller segments are considered as vertices of the approximating polygon. Experimental results show that the proposed method is robust and efficient. Keywords: Polygonal approximation, Split approach, Eigenvalue, Integral square error, Compression ratio. [ABSTRACT FROM AUTHOR]

Recognition of voice in a multi-speaker environment involves speech separation, speech feature extraction and speech feature matching. Though traditionally vector quantization is one of the algorithms used for speaker recognition; its effectiveness is not well appreciated in case of noisy or multi-speaker environment. This paper describes the usability of the Independent Component Analysis (ICA) technique to enhance the effectiveness of speaker recognition using vector quantization. Results obtained by this approach are compared with that obtained using a more direct approach to establish the usefulness of the proposed method. Keywords: Speech recognition, ICA, MFCC, Vector Quantization. [ABSTRACT FROM AUTHOR]

The problem of detecting anomalous network connections caused by intrusion activities is called Network intrusion detection. Conventional classification methods use data from both normal and intrusion classes to build the classifiers. However, intrusion data are usually scarce and difficult to collect. Novelty detection approach overcomes this problem which depends only on normal data. For this purpose, nonparametric density estimation approaches based on Parzen-window estimators are proposed earlier. Two fundamental problems faced are, (i) due to curse of dimensionality, for high dimensional data with a limited training set, the estimation can be biased and (ii) high computational requirements. We propose, (i) a novel pattern synthesis technique to synthesize artificial new training patterns to increase the training set size and thus to reduce the curse of dimensionality effect, and (ii) a compact data representation scheme to store the entire synthetic set to reduce the computational costs. The effectiveness of our methods are experimentally demonstrated. [ABSTRACT FROM AUTHOR]

Segmentation of magnetic resonance imaging (MRI) data entails assigning tissue class labels to voxels. The primary source of segmentation error is the partial volume effect (PVE) which occurs most often with low resolution imaging - With large voxels, the probability of a voxel containing multiple tissue classes increases. Although the PVE problem has not been solved, the first stage entails correctly identifying PVE voxels. We employ rough sets to identify them automatically. [ABSTRACT FROM AUTHOR]

Textual case bases can contain thousands of features in the form of tokens or words, which can inhibit classification performance. Recent developments in rough set theory and its applications to feature selection offer promising approaches for selecting and reducing the number of features. We adapt two rough set feature selection methods for use on n-ary class text categorization problems. We also introduce a new method for selecting features that computes the union of features selected from randomly-partitioned training subsets. Our comparative evaluation of our method with a conventional method on the Reuters-21578 data set shows that it can dramatically decrease training time without compromising classification accuracy. Also, we found that randomized training set partitions dramatically reduce training time. [ABSTRACT FROM AUTHOR]

Approximation spaces are fundamental for the rough set approach. We discuss their application in machine learning and pattern recognition. Keywords: Rough sets, approximation spaces, concept approximation. [ABSTRACT FROM AUTHOR]

We discuss some rough set tools for perception modelling that have been developed in our project for a system for modelling networks of classifiers for compound concepts. Such networks make it possible to recognize behavioral patterns of objects and their parts changing over time. We present a method that we call a method for on-line elimination of non-relevant parts (ENP). This method was developed for on-line elimination of complex object parts that are irrelevant for identifying a given behavioral pattern. Some results of experiments with data from the road simulator are included. [ABSTRACT FROM AUTHOR]

A genetic algorithm (GA) based multiple route selection for car navigation device is proposed in this work. The proposed scheme offers the driver a choice from alternate near optimal solutions, each of which carries some specific characteristics based on the knowledge of the road map and the environment. The algorithm has been simulated on some real road map and it is found that it provides better solution compared to deterministic algorithms and other GA based algorithms in terms of driver's satisfaction. [ABSTRACT FROM AUTHOR]

The strongly NP-complete Double Digest Problem (DDP) for physical mapping of DNA, is now used for efficient genotyping. An instance of DDP has multiple distinct solutions. Existing methods produce a single solution, and are slow for large instances. We employ a type of equivalence among the distinct solutions to obtain almost all of them. Our method comprises of first finding a solution from each equivalence class by an elitist genetic algorithm, and then generating entire classes. Notable efficiency was achieved due to significant reduction in search space. [ABSTRACT FROM AUTHOR]

The main obstacle in realizing semantic-based image retrieval is from the web that semantic description of an image is difficult to capture in low-level features. Text based keywords can be generated from web documents to capture semantic information narrowing down the search space. We use an effective dynamic approach to integrate keywords and color-texture features to take advantage of their complementing strengths. Experimental results show that the integrated approach has better retrieval performance than both the text based and the content-based techniques. [ABSTRACT FROM AUTHOR]

We address the problem of fusing sparse and noisy depth data obtained from a range finder with features obtained from intensity images to estimate ego-motion and refine 3D structure of a scene using a Rao-Blackwellized particle filter. For scenes with low depth variability, the algorithm shows an alternate way of performing Structure from Motion (SfM) starting with a flat depth map. Instead of using 3D depths, we formulate the problem using 2D image domain parallax and show that conditioned on non-linear motion parameters, the parallax magnitude with respect to the projection of the vanishing point forms a linear subsystem independent of camera motion and their distributions can be analytically integrated. Thus, the structure is obtained by estimating parallax with respect to the given depths using a Kalman filter and only the ego-motion is estimated using a particle filter. Hence, the required number of particles becomes independent of the number of feature points which is an improvement over previous algorithms. Experimental results on both synthetic and real data show the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

The Discrete Hodge Helmholtz Decomposition (DHHD) is able to locate critical points in a vector field. We explore two novel applications of this technique to image processing problems, viz., hurricane tracking and fingerprint analysis. The eye of the hurricane represents a rotational center, which is shown to be robustly detected using DHHD. This is followed by an automatic segmentation and tracking of the hurricane eye, which does not require manual initializations. DHHD is also used for identification of reference points in fingerprints. The new technique for reference point detection is relatively insensitive to noise in the orientation field. The DHHD based method is shown to detect reference points correctly for 96.25% of the images in the database used. [ABSTRACT FROM AUTHOR]

We present a novel approach for extracting moving objects, suitable for intrusion detection and video surveillance systems. Proposed method is characterized by robustness to illumination changes, acclimation to the changes in constituents of background and significantly reduced false alarm rate. We extract pieces of edge information from images and represent these segments with efficiently designed edge classes. Proposed algorithm for matching and updating of edges incorporates the robustness and resilience to intrusion detection system, which is illustrated by the results of our experiments. [ABSTRACT FROM AUTHOR]

A hierarchical framework for document segmentation is proposed as an optimization problem. The model incorporates the dependencies between various levels of the hierarchy unlike traditional document segmentation algorithms. This framework is applied to learn the parameters of the document segmentation algorithm using optimization methods like gradient descent and Q-learning. The novelty of our approach lies in learning the segmentation parameters in the absence of groundtruth. [ABSTRACT FROM AUTHOR]

The non-linearity exhibited by the non-classical receptive field in human visual system has been combined with the linear classical receptive field model. This enables us to construct higher order Gaussian Derivatives as a linear combination of lower order derivatives at different scales. Based on this, a new kernel which simulates non-classical receptive fields with extended disinhibitory surrounds, has been proposed. It is easy to implement and finds justification from an old psychophysical angle too. The proposed kernel has been shown to perform better than the well-known Laplacian kernel, which models the classical excitatory-inhibitory receptive fields. [ABSTRACT FROM AUTHOR]

The postal address data and the domain information for address validation contain qualitative, numeric, interval and other types of data. The efficient processing of such data required for postal automation needs a robust data structure that facilitates their storage and access. A symbolic data structure is proposed to represent the postal address and the information relevant for validating the postal address is stored in a newly devised symbolic knowledge base. The symbolic representation gives a formal structure to the information and hence is more beneficial than other representations such as frames, which do not reflect the structure inherent in the domain knowledge. The process of postal address validation checks the different components of the postal address for consistency before using it for further processing. In the present work a symbolic knowledge base supported address validation system is developed and tested for about 500 addresses. The system efficiency is observed to be 95.6% in validating the addresses automatically. Keywords: Postal Address validation, Symbolic object, knowledge base, Frames. [ABSTRACT FROM AUTHOR]

A method for segmentation of magnetic resonance (MR) images of the human brain using a fuzzy adaptive radial basis function neural network (FARBF-NN) has been proposed. Since the quality of MR images always gets affected by intensity in-homogeneities (artifacts or noises), generated due to the non-uniformity of magnetic fields during the acquisition process, thereby making segmentation task more difficult. The outputs of the hidden layer neurons of the FARBF-NN have been modified using a fuzzy membership function to eliminate the effect of noises present in the input image. The proposed method has been tested both on simulated and real patient MR brain images for segmentation and found to be better than the k-means clustering algorithm, the fuzzy c-means (FCM) clustering algorithm, and the RBF neural network that uses k-means clustering algorithm to select the centers of the RBFs in the hidden layer, in most of the cases. [ABSTRACT FROM AUTHOR]

An algorithm for fusing multiple remotely sensed image classifiers is addressed herein using fuzzy integral with error proportionate fuzzy measures. This method includes a procedure for calculating the λ-fuzzy measures which are adjusted depending on error correlation among the individual classifiers. Based on these fuzzy measures, the fuzzy integral is then used as non-linear function to search for maximum degree of agreement between multiple conflicting sources of evidence. Results obtained are used for decision making in classification problem. Experimental results on classification of remotely sensed images show that the performance of proposed multi-classifier method performs better than conventional method where fixed fuzzy measures are used. [ABSTRACT FROM AUTHOR]

A recent study on symbolic data analysis literature reveals that symbolic distance measures are playing a major role in solving the pattern recognition and analysis problems. After a careful study on the existing symbolic distance measures, we have identified that most of the existing symbolic distance measures either suffer from generalization or do not address object variability. To alleviate these problems we are proposing new generalized Similarity symbolic distance measure. The proposed distance measure is asymmetric, addresses object variability, and obeys partial order. To leverage the advantages of both fuzzy set theory and symbolic data analysis, conventional classification algorithm that works on the principles of fuzzy equivalence relation has been extended to handle Symbolic data. Efficacies of the proposed techniques are validated by conducting several experiments on the well-known assertion type of symbolic data sets with known classification results. Keywords: Fuzzy-Symbolic data analysis, Fuzzy hierarchical analysis, Symbolic distance measures. [ABSTRACT FROM AUTHOR]

A Human-Computer Interaction (HCI) system has been developed with an Artificial Neural Network (ANN) using a motion tracker and a data glove. The HCI system is able to recognize American Sign Language letter and number gestures. The finger joint angle data obtained from the strain gauges in the sensory glove define the hand shape while the data from the motion tracker describe the hand position and orientation. The data flow from the sensory glove is controlled by a software trigger using the data from the motion tracker during signing. Then, the glove data is processed by a recognition neural network. [ABSTRACT FROM AUTHOR]

We propose a novel hybrid Hölder-SVM detection algorithm for arrhythmia classification. The Hölder exponents are computed efficiently using the wavelet transform modulus maxima (WTMM) method. The hybrid system performance is evaluated using the benchmark MIT-BIH arrhythmia database. The implemented model classifies 160 of Normal sinus rhythm, 25 of Ventricular bigeminy, 155 of Atrial fibrillation and 146 of Nodal (A-V junctional) rhythm with 96.94% accuracy. The distinct scaling properties of different types of heart rhythms may be of clinical importance. [ABSTRACT FROM AUTHOR]

An algorithm of PCA face recognition based on topological manifolds theory is proposed, which based on the sample sets' topological character in the feature space which is different from "classification". Compare with the traditional PCA+ NN algorithm, experiments prove its efficiency. [ABSTRACT FROM AUTHOR]

The work proposes a hierarchical architecture for learning from dynamic scenes at various levels of knowledge abstraction. The raw visual information is processed at different stages to generate hybrid symbolic/sub-symbolic descriptions of the scene, agents and events. The background is incrementally learned at the lowest layer, which is used further in the mid-level for multi-agent tracking with symbolic reasoning. The agent/event discovery is performed at the next higher layer by processing the agent features, status history and trajectory. Unlike existing vision systems, the proposed algorithm does not assume any prior information and aims at learning the scene/agent/event models from the acquired images. This makes it a versatile vision system capable of performing in a wide variety of environments. [ABSTRACT FROM AUTHOR]

In this study, we present a new multi-band image representation for improving AAM segmentation accuracy for illumination invariant face alignment. AAM is known to be very sensitive to the illumination variations. We have shown that edges, originating from object boundaries are far less susceptible to illumination changes. Here, we propose a contour selector which mostly collects contours originating from boundaries of the face components (eyes, nose, chin, etc.) and eliminates the others arising from texture. Rather than representing the image using grey values, we use Hill, Hue and Grey value (HHG) for image representation. We demonstrate that HHG representation gives more accurate and reliable results as compared to image intensity alone under various lighting conditions. [ABSTRACT FROM AUTHOR]