General Pattern Recognition Research Articles

Star Pattern Identification Technique by Modified Grid Algorithm

A star pattern identification algorithm, called modified grid algorithm, for attitude determination of spacecraft is addressed. The proposed algorithm is closely connected to a general pattern recognition technique. Each star is characterized by a well-defined pattern that can be determined by the surrounding stars. The so-called grid algorithm is one of the star identification algorithms based upon the pattern recognition approach. The proposed algorithm, is motivated by the conventional grid algorithm. To enhance the performance of the conventional grid algorithm, a modified method using polar grid, virtual grid, and multi-references is proposed. Simulation study is conducted for the demonstration of the new algorithm. The proposed modified grid approach turns out to make the grid algorithm more robust and reliable.

Discriminative training via minimization of risk estimates based on Parzen smoothing

We describe a new approach to estimating classification risk in the domain of a suitably defined transformation that can be used as the basis for optimization of generic pattern recognition systems, including hidden Markov models and Multi-Layer Perceptrons. The two formulations of risk estimate described here are closely tied to the Minimum Classification Error/Generalized Probabilistic Descent (MCE/GPD) framework for discriminative training that is well-known to the speech recognition community. In the new approach, high-dimensional and possibly variable-length training tokens are mapped to the centers of Parzen kernels which are then easily integrated to find the risk estimate. The utility of such risk estimates lies in the fact that they are explicit functions of the system parameters and hence suitable for use in practical optimization methods. The use of Parzen estimation makes it possible to establish convergence of the risk estimate to the true theoretical classification risk, a result that formally expresses the benefit of linking the degree of smoothing to the training set size. Convergence of the minimized risk estimate is also analyzed. The new approach establishes a more general theoretical foundation for discriminative training than existed before, supporting previous work and suggesting new variations for future work.

Toll-like receptor recognition of Toxoplasma gondii

Toxoplasma gondii potently stimulates IFN-γ production by both the innate and adaptive immune system as part of its host adaptation. This response is known to be dependent on an Myeloid Differentiation factor 88 signaling pathway used by Toll-like receptors (TLRs), a family of proteins involved in the recognition of microbial molecular patterns. In the following review, we summarise the evidence for specific TLR function in host resistance to T. gondii focusing on the recent discovery in the parasite of a profilin-like ligand that potently stimulates TLR11 and regulates the production of IL-12, a cytokine necessary for the protective IFN-γ response. In addition, we discuss the hypothesis that TLR11 may have evolved as a general pattern recognition receptor for apicomplexan protozoa and that as highly conserved proteins associated with actin-based motility, profilins are logical ligand targets for this form of pathogen detection. Finally, we review the evidence for involvement of other TLR and TLR ligands in host resistance to T. gondii and discuss how such receptors might synergise with TLR11 in the innate response to the parasite.

Pattern recognition methods for advanced stochastic protein sequence analysis using HMMs

Currently, Profile Hidden Markov Models (Profile HMMs) are the methodology of choice for probabilistic protein family modeling. Unfortunately, despite substantial progress the general problem of remote homology analysis is still far from being solved. In this article we propose new approaches for robust protein family modeling by consequently exploiting general pattern recognition techniques. A new feature based representation of amino acid sequences serves as the basis for semi-continuous protein family HMMs. Due to this paradigm shift in processing biological sequences the complexity of family models can be reduced substantially resulting in less parameters which need to be trained. This is especially favorable when only little training data is available as in most current tasks of molecular biology research. In various experiments we prove the superior performance of advanced stochastic protein family modeling for remote homology analysis which is especially relevant for e.g. drug discovery applications.

A highly robust estimator for regression models

It is well known that classical robust estimators tolerate only less than fifty percent of outliers. However, situations with more than fifty percent of outliers often occur in practice. The efficient identification of objects from a noisier background is thus a difficult problem. In this paper, a highly robust estimator is formulated to tackle such a difficulty. The proposed estimator is called the regression density decomposition (RDD) estimator. The computational analysis of the estimator and its properties are discussed and a simulated annealing algorithm is proposed for its implementation. It is demonstrated that the RDD estimator can resist a very large proportion of noisy data, even more than fifty percent. It is successfully applied to some simulated and real-life noisy data sets. It appears that the estimator can solve efficiently and effectively general regression problems, pattern recognition, computer vision and data mining problems.

OCR in Indian Scripts: A Survey

India is a multi-lingual country. A significantly large number of scripts are used to represent these languages. A desire of vision researchers is to develop an integrated Optical Character Recognition (OCR) system which will be able to process all such scripts. Such a development, if objectified, will not only enable faster flow of information across the country, but also have a profound impact on its scientific and economic development. Courageous endeavors have been successfully made towards the development of a system capable of recognizing machine-printed, or hand-written characters and/or numerals. However, most Indian scripts do not have an integrated OCR system. Further the development of a unified system which is capable of processing all Indian scripts is still a dream. This article presents a survey of the current literature on the development of OCR's in Indian scripts. Reviewing the basics of and the motivation towards the development of OCR system, the article analyzes the various methodologies employed in general purpose pattern recognition system. A critical analysis of the work towards OCR system in Indian languages, with pointers towards possible future work is also presented.

Social Network Schemas and the Learning of Incomplete Networks.

Social networks that are missing relations among some of their members--termed incomplete networks--have been of critical theoretical and empirical interest in sociological research on weak ties and structural holes but typically have been overlooked in social psychological studies of network learning. Five studies tested for schematic processing differences in the encoding and recalling of incomplete networks. In Studies 1 and 2, prior knowledge of missing relations facilitated learning an unfamiliar, incomplete network. Study 3 ruled out differences in general pattern recognition ability as an explanation. Study 4 manipulated the degree of familiarity with missing relations, which produced predicted differences in learning rates. Finally, Study 5 examined how improved learning of an incomplete network affected a strategic organizational choice. The findings suggest that people can become schematic for complex, incomplete social networks.

TJ-II wave forms analysis with wavelets and support vector machines

Since the fusion plasma experiment generates hundreds of signals, it is essential to have automatic mechanisms for searching similarities and retrieving of specific data in the wave form database. Wavelet transform (WT) is a transformation that allows one to map signals to spaces of lower dimensionality. Support vector machine (SVM) is a very effective method for general purpose pattern recognition. Given a set of input vectors which belong to two different classes, the SVM maps the inputs into a high-dimensional feature space through some nonlinear mapping, where an optimal separating hyperplane is constructed. In this work, the combined use of WT and SVM is proposed for searching and retrieving similar wave forms in the TJ-II database. In a first stage, plasma signals will be preprocessed by WT to reduce their dimensionality and to extract their main features. In the next stage, and using the smoothed signals produced by the WT, SVM will be applied to show up the efficiency of the proposed method to deal with the problem of sorting out thousands of fusion plasma signals. From observation of several experiments, our WT+SVM method is very viable, and the results seems promising. However, we have further work to do. We have to finish the development of a Matlab toolbox for WT+SVM processing and to include new relevant features in the SVM inputs to improve the technique. We have also to make a better preprocessing of the input signals and to study the performance of other generic and self custom kernels. To reach it, and since the preprocessing stages are very time consuming, we are going to study the viability of using DSPs, RPGAs or parallel programming techniques to reduce the execution time.

A neural multiclassifier system for object recognition in robotic vision applications

A robotic vision system has been developed that is based on the optical recognition of objects to be manipulated, that are located in the workspace of the robotic manipulator. The developed system has a low development and operation cost, is controlled via an external computer and operates in an unstructured complex environment. The vision system is desired to recognize the objects, which are placed in the workspace and also to identify the exact position and orientation of each particular object, in order to lead the robot manipulator system. For the recognition of objects, a high performance NEural MUlticlassifier System (NEMUS) is presented, which combines multiple classifiers that operate on different feature sets. NEMUS is characterized by a great degree of modularity and flexibility and is very efficient for demanding and generic pattern recognition applications.

Automatic detection of welding defects using texture features

In this paper we present a new approach to detecting weld defects from digitalised films based on texture features. Texture is one of the most important features used in recognising patterns in an image. However, these features are not yet commonly exploited in the analysis of X-ray images in NDT. The paper describes two groups of widely used texture features: 1) features based on the cooccurrence matrix, which gives a measurement of how often one grey value will appear in a specified spatial relationship to another grey value on the image; and 2) features based on 2D Gabor functions, i.e., Gaussian shaped band-pass filters, with dyadic treatment of the radial spatial frequency range and multiple orientations, which represent an appropriate choice for tasks requiring simultaneous measurement in both space and frequency domains. The proposed approach to detecting weld defects follows a general pattern recognition scheme based on three steps: segmentation, feature extraction and classification. That is, in our case, 1) potential defects are segmented using an edge detector based on the Laplacian-of-Gauss operator; 2) texture features of the potential defects are extracted; and 3) the most relevant features are used as input data on a statistical classifier. This preliminary study makes a contribution to the improvement of the automatic detection of welding defects.

Prediction of biological activity for high-throughput screening using binary kernel discrimination.

High-throughput screening has made a significant impact on drug discovery, but there is an acknowledged need for quantitative methods to analyze screening results and predict the activity of further compounds. In this paper we introduce one such method, binary kernel discrimination, and investigate its performance on two datasets; the first is a set of 1650 monoamine oxidase inhibitors, and the second a set of 101 437 compounds from an in-house enzyme assay. We compare the performance of binary kernel discrimination with a simple procedure which we call "merged similarity search", and also with a feedforward neural network. Binary kernel discrimination is shown to perform robustly with varying quantities of training data and also in the presence of noisy data. We conclude by highlighting the importance of the judicious use of general pattern recognition techniques for compound selection.

Characterisation of proteinaceous binders and drying oils in wall painting samples by gas chromatography–mass spectrometry

A reliable analytical procedure has been developed for the characterisation of drying oils and proteinaceous binders in samples of painted artworks. The method is based on microwave assisted acid hydrolysis followed by the extraction of the lipid component with diethyl ether. Fatty acids were determined after saponification of the organic phase and derivatisation with N- tert.-Butyl methylsilyl- N-methyltrifluoroacetamide. Amino acids present in the acid aqueous phase were derivatised with the same silyl agent. The separation and quantitative determination of derivatives was performed by gas chromatography–mass spectrometry. Detection limits of fatty acids (lauric, myristic, palmitic, stearic, oleic, suberic, azelaic and sebacic acid) and amino acids were in the range 50–100 pg. The mean recovery of the procedure for lipid determination was about 70% and that for proteins 100% with a reproducibility better than 10% for 1 mg samples. The overall procedure was validated by analysing suitable reference wall painting samples. The identification of the proteinaceous binders was achieved by principal component analysis performed on the fourteen amino acid relative percentages while the identification of the drying oils was based on the values of palmitic to stearic ratio and azelaic to palmitic ratio. A general pattern recognition scheme based on these values and on the amounts of dicarboxylic acids and the presence of cholesterol was developed to distinguish between proteinaceous binder, “tempera grassa” and drying oils. The application of the proposed procedure to samples from “ The Legend of True Cross” (1452) by Piero della Francesca allowed the identification of “tempera grassa” made of egg and linseed oil.

Application of Artificial Neural Networks to Medical Image Pattern Recognition: Detection of Clustered Microcalcifications on Mammograms and Lung Cancer on Chest Radiographs

Three neural network models were employed to evaluate their performances in the recognition of medical image patterns associated with lung cancer and breast cancer in radiography. The first method was a pattern match neural network. The second was a conventional backpropagation neural network. The third method was a backpropagation trained neocognitron in which the signal propagation is operated with the convolution calculation from one layer to the next. In the convolution neural network (CNN) experiment, several output association methods and trainer imposed driving functions in conjunction with the convolution neural network are proposed for general medical image pattern recognition. An unconventional method of applying rotation and shift invariance is also used to enhance the performance of the neural nets. We have tested these methods for the detection of microcalcifications on mammograms and lung nodules on chest radiographs. Pre-scan methods were previously described in our early publications. The artificial neural networks act as final detection classifiers to determine if a disease pattern is presented on the suspected image area. We found that the convolution neural network, which internally performs feature extraction and classification, achieves the best performance among the three neural network models. These results show that some processing associated with disease feature extraction is a necessary step before a classifier can make an accurate determination.

Translation-, rotation-, scale-, and distortion-invariant object recognition through self-organization.

The task of visual object recognition is often complicated by the fact that a single 3-D object can undergo a number of transformations which can substantially alter its projection onto a 2-D surface, such as the retina. Such transformations include translation of the object in the visual field, changes in the size of the object, its orientation in the 2-D plane and the viewing perspective. For a general pattern recognition system to detect and recognize and object after such transformations, it must be able to associate widely differing patterns with the same object label. In this paper, a novel self-organizing model, called the Multiple Elastic Modules (MEM), is presented which attempts to solve this problem by searching a multi-dimensional space, where each axis is defined by one of the transformations (e.g. scale, translation, rotation, etc.). A particular object of a specific size, orientation and spatial location is mapped onto a single point in this space. Of course, distortions and minor variations in an object's image will expand this point to a small localized area in this multi-dimensional space. Such a powerful representation scheme comes at a cost of high computational demand due to the combinatorially large search space. The MEM approach to solving this problem efficiently partitions the solution space to search the most promising areas for the correct match. Simulation results are presented on detecting a stick-figure object under translation, distortion, scale, and rotation transformations in a cluttered background.

Artificial convolution neural network for medical image pattern recognition

We have developed several training methods in conjunction with a convolution neural network for general medical image pattern recognition. An unconventional method of using rotation and shift invariance is also proposed to enhance the neural net performance. The structure of the artificial neural network is a simplified network structure of the neocognitron. Two-dimensional local connection as a group is the fundamental architecture for the signal propagation in the convolution neural network. Weighting coefficients of convolution kernels are formed by the neural network through backpropagated training for this artificial neural net. In addition, radiologists' reading procedure was modelled in order to instruct the artificial neural network to recognize the predefined image patterns and those of interest to experts. Our training techniques involve (a) radiologists' rating for each suspected image area, (b) backpropagation of generalized distribution, (c) trainer imposed functions, (d) shift and rotation invariance of diagnosis interpretation, and (e) consistency of clinical input data using appropriate background reduction functions. We have tested these methods for detecting lung nodules on chest radiographs and microcalcications on mammograms. The performance studies have shown the potential use of this technique in a clinical environment. We also used a profile double-matching technique for initial nodule search and used a wavelet high pass filtering technique to enhance subtle clustered microcalcifications. We set searching parameters at a highly sensitive level to identify all potential disease areas. The artificial convolution neural network acts as a final detection classifier to determine whether a disease pattern is shown on the suspected image area.

Common theoretical formulation of the pattern recognition, identification, and detection problems

Patterns are formalized as operators, which may be compared on the basis of an equivalence relation, of a metricd, and of probability measures. The general pattern recognition problem for metric and probabilistic patterns is then formulated, referring to the formalism of U. Grenander; identification and detection are shown to be special cases hereof. A list of possible distancesd between patterns is then given, including a suggested distance between line patterns; wider applications of some of these distances are suggested. This paper has a theoretical concern, but it nevertheless aims at letting pattern recognition draw benefit from techniques which are similar in theory, but which are curiously still considered as distinct in practice.

Toward robust handwritten character recognition

This paper addresses the problem of establishing a robust methodology for handwritten character recognition. First, conventional techniques in Japan are critically surveyed to clarify the state of the art and remaining problems. Second, new and promising approaches are enumerated from the viewpoint of general pattern recognition methodology. Third, a challenge to distortion-tolerant shape matching using the idea of local affine transformation is described. Finally, interesting research topics relevant to mailing address recognition are discussed.

Is an Object an Object an Object? Cognitive and Neuropsychological Investigations of Domain Specificity in Visual Object Recognition

Are all types of objects recog nized in the same way, or are differ ent kinds of visual object recognition systems used to recognize different types of objects? Most current work on object recognition in cognitive science has assumed, explicitly or implicitly, that all visual stimuli are recognized by a common set of mechanisms. Cognitive scientists such as Marr1 and Biederman,2 who have proposed comprehensive theo ries of object recognition, do not specify different types of representa tions or processes for different types of stimuli. Rather, these scientists have described a single type of sys tem capable of recognizing as wide a range of stimuli as possible. Other researchers have ques tioned the existence of a general purpose pattern recognition system, and have instead suggested that the visual system has evolved numerous specialized subsystems for recogniz ing different types of stimuli. Proba bly the most extreme proponent of this view was Konorski,3 who sug gested that there were nine different subsystems used in visual recogni

Satellite Monitoring of the Earth

Part 1 Selected platforms and sensors: general considerations satellite platforms satellite remote sensors. Part 2 Atmospheric considerations: interaction between the atmosphere and radiation global distribution of atmospheric water ozone distribution. Part 3 Spectral characteristics of natural systems: general considerations properties of water spectral characteristics of geological targets spectral characteristics of vegetation apparent radiance. Part 4 Concepts in data interpretation: general considerations spectral bands and ground resolution pattern recognition analog enhancement digital image processing interpretation of radar images. Part 5 Observations over the oceans: general considerations currents and upwelling systems. Part 6 Convergence zones: subtropical and antarctic convergence zones bathymetry sea ice sea state, wind speed and direction sea surface topography. Part 7 Observations over the continents: geological features water on the continent vegetation.

Generalized guard-zone algorithm (GGA) for learning: automatic selection of threshold

This work is a continuation of our earlier work on the Generalized Guard-zones Algorithm (GGA) for self-supervised parameter learning. An attempt is made here for the automatic determination of the guard-zone parameter λ n (i.e. the threshold used for discarding doubtful or mislabeled samples) at every instant of learning, for the general m-class N-feature pattern recognition problem. This is done by minimizing the mean squared error (MSE) of the estimate, under a simple probabilistic model which takes into consideration the presence of mislabeled training samples. Under the assumptions of normality, it is found that the estimates for λ n , so obtained are distribution-free, that is, they do not depend on the parameters of the distribution. They are functions of N, the iteration number n and certain percentage points of the beta distribution with parameters N and n - N. The effectiveness of the automatic selection of guard-zone dimension is further demonstrated on a bivariate three-class data set to show the improvement in performance of the GGA.