Two-class Pattern Recognition Problem Research Articles

Self-supervised deep reconstruction of mixed strip-shredded text documents

The reconstruction of shredded documents consists of coherently arranging fragments of paper (shreds) to recover the original document(s). A great challenge in computational reconstruction is to properly evaluate the compatibility between the shreds. While traditional pixel-based approaches are not robust to real shredding, more sophisticated solutions compromise significantly time performance. The solution presented in this work extends our previous deep learning method for single-page reconstruction to a more realistic/complex scenario: the reconstruction of several mixed shredded documents at once. In our approach, the compatibility evaluation is modeled as a two-class (valid or invalid) pattern recognition problem. The model is trained in a self-supervised manner on samples extracted from simulated-shredded documents, which obviates manual annotation. Experimental results on three datasets – including a new collection of 100 strip-shredded documents produced for this work – have shown that the proposed method outperforms the competing ones on complex scenarios, achieving accuracy superior to 90%.

Detection of vocal fold paralysis and oedema using time-domain features and Probabilistic Neural Network

This paper proposes a feature extraction method based on time-domain energy variation for the detection of vocal fold pathology. In this work, two different vocal fold problems (vocal fold paralysis and edema) are taken for analysis and in either case, a two-class pattern recognition problem is investigated. The normal and pathological speech samples are used from Massachusetts Eye and Ear Infirmary database. Probabilistic Neural Network (PNN) is employed for the classification. The experimental results show that the proposed features give very promising classification accuracy of 90% and can be used to detect the vocal fold paralysis and edema clinically.

Blind detection of image splicing based on image quality metrics and moment features

Image splicing is a technique commonly used in image tampering. To implement image splicing blind detection,a new splicing detection scheme was proposed. Image splicing detection could be regarded as a two-class pattern recognition problem and the model was established based on moment features and some Image Quality Metrics (IQMs) extracted from the given test image. This model could measure statistical differences between original image and spliced image. Kernel-based Support Vector Machine (SVM) was chosen as a classifier to train and test the given images. Experimental results demonstrate that this new splicing detection scheme has some advantages of high-accuracy and wide-application.

Machine learning multi-classifiers for peptide classification

In this paper, we study the performance improvement that it is possible to obtain combining classifiers based on different notions (each trained using a different physicochemical property of amino-acids). This multi-classifier has been tested in three problems: HIV-protease; recognition of T-cell epitopes; predictive vaccinology. We propose a multi-classifier that combines a classifier that approaches the problem as a two-class pattern recognition problem and a method based on a one-class classifier. Several classifiers combined with the “sum rule” enables us to obtain an improvement performance over the best results previously published in the literature.

Advanced methods for two-class pattern recognition problem formulation for minutiae-based fingerprint verification

We present a new method for minutiae-based fingerprint verification that approaches the problem as a two-class pattern recognition problem. In our knowledge, this is one of the first works that uses as features for fingerprint verification the response of a minutiae matcher between two fingerprint. The feature vector obtained by the minutiae matching is classified into “genuine” or “impostor” by support vector machines. Results from FVC2002 are presented, yielding remarkable performance improvement with respect to other state-of-the-art approaches.

Advanced methods for two-class problem formulation for on-line signature verification

We present several systems for on-line signature verification that approach the problem as a two-class pattern recognition problem. To our knowledge, this is the first work that solves the problem of on-line signature verification as a two-class problem using global (and not local) features. The feature vector obtained by global features is then classified into one of the two classes (genuine or impostor) by a support vector machine. Moreover, we show the combination of the systems introduced in this work permit a dramatic reduction of the equal error rate.

A novel method for fingerprint verification that approaches the problem as a two-class pattern recognition problem

We present a system for fingerprint verification that approaches the problem as a two-class pattern recognition problem. The distances of the test fingerprint to the reference fingerprints are normalized by the corresponding mean values obtained from the reference set, to form a five-dimensional feature vector. This feature vector is then projected onto a one-dimensional Karhunen-Loeve space and then classified into one of the two classes (genuine or impostor).

Two-class fingerprint matcher

We present a system for fingerprint verification that approaches the problem as a two-class pattern recognition problem. The features extracted by “FingerCode” are used to capture the ridge strength. This feature vector is then classified as genuine or impostor according to a novel approach to handle the fingerprint verification as a two-class problem. Moreover, we show that extracting the features from sub-images around the core permits to better represent the local information.

Identity authentication using improved online signature verification method

We present a system for online handwritten signature verification, approaching the problem as a two-class pattern recognition problem. A test signature’s authenticity is established by first aligning it with each reference signature for the claimed user, using dynamic time warping. The distances of the test signature to the nearest, farthest and template reference signatures are normalized by the corresponding mean values obtained from the reference set, to form a three-dimensional feature vector. This feature vector is then classified into one of the two classes (genuine or forgery). A linear classifier used in conjunction with the principal component analysis obtained a 1.4% error rate for a data set of 94 people and 619 test signatures (genuine signatures and skilled forgeries). Our method received the first place at SVC2004 with a 2.8% error rate.

Stability regions for constrained nonlinear systems and their functional characterization via support-vector-machine learning

This paper develops a computational approach for characterizing the stability regions of constrained nonlinear systems. A decision function is constructed that allows arbitrary initial states to be queried for inclusion within the stability region. Data essential to the construction process are generated by simulating the nonlinear system with multiple initial states. Using special procedures based on known properties of the stability region, the state data are randomly selected so that they are concentrated in desirable locations near the boundary of the stability region. Selected states belong either to the stability region or do not, thus producing a two-class pattern recognition problem. Support vector machine learning, applied to this problem, determines the decision function. Special techniques are introduced that significantly improve the accuracy and efficiency of the learning process. Numerical examples illustrate the effectiveness of the overall approach.

A novel featureless approach to mass detection in digital mammograms based on support vector machines

In this work, we present a novel approach to mass detection in digital mammograms. The great variability of the appearance of masses is the main obstacle to building a mass detection method. It is indeed demanding to characterize all the varieties of masses with a reduced set of features. Hence, in our approach we have chosen not to extract any feature, for the detection of the region of interest; in contrast, we exploit all the information available on the image. A multiresolution overcomplete wavelet representation is performed, in order to codify the image with redundancy of information. The vectors of the very-large space obtained are then provided to a first support vector machine (SVM) classifier. The detection task is considered here as a two-class pattern recognition problem: crops are classified as suspect or not, by using this SVM classifier. False candidates are eliminated with a second cascaded SVM. To further reduce the number of false positives, an ensemble of experts is applied: the final suspect regions are achieved by using a voting strategy. The sensitivity of the presented system is nearly 80% with a false-positive rate of 1.1 marks per image, estimated on images coming from the USF DDSM database.

Reduction in correlation-filter sensitivity to background clutter by the automatic spatial frequency selection algorithm

We propose an automatic spatial frequency selection correlation filter that reduces the sensitivity to nonoverlapping noise or background clutter. This is achieved by inclusion of distorted versions of the reference images surrounded by nonoverlapping background clutter. Furthermore, we impose that the window functions of the reference images give response zero-correlation amplitudes. Simulation results are provided in the case of a two-class pattern-recognition problem and show that the results are appreciably increased. The results are compared with a normal automatic spatial frequency selection.

A new adaptive polynomial neural network

This paper considers the problem of the construction of nonlinear mapping by using an adaptive polynomial neural network [1], implementing a learning rule. First we apply the method to a two-class pattern recognition problem. We use one high order neuron with a threshold element ranging from −1 to +1. Positive output means class 1 and negative output means class 2. The main idea of the method proposed is that the weights are adjusted automatically in such a way to move the decision boundary to a place of low pattern density. Once reached the convergence, to improve the generalization ability, we add a growing noise to the data available. The training is performed by a steepest-descent algorithm on the weights.

On the finite sample performance of the nearest neighbor classifier

The finite sample performance of a nearest neighbor classifier is analyzed for a two-class pattern recognition problem. An exact integral expression is derived for the m-sample risk R/sub m/ given that a reference m-sample of labeled points is available to the classifier. The statistical setup assumes that the pattern classes arise in nature with fixed a priori probabilities and that points representing the classes are drawn from Euclidean n-space according to fixed class-conditional probability distributions. The sample is assumed to consist of m independently generated class-labeled points. For a family of smooth class-conditional distributions characterized by asymptotic expansions in general form, it is shown that the m-sample risk R/sub m/ has a complete asymptotic series expansion R/sub m//spl sim/R/sub /spl infin//+/spl Sigma//sub k=2//sup /spl infin//c/sub k/m/sup -k/n/ (m/spl rarr//spl infin/), where R/sub /spl infin// denotes the nearest neighbor risk in the infinite-sample limit and the coefficients c/sub k/ are distribution-dependent constants independent of the sample size m. The analysis thus provides further analytic validation of Bellman's curse of dimensionality. Numerical simulations corroborating the formal results are included, and extensions of the theory discussed. The analysis also contains a novel application of Laplace's asymptotic method of integration to a multidimensional integral where the integrand attains its maximum on a continuum of points. >

An unsupervised learning technique for artificial neural networks

A new artificial neural model for unsupervised learning is proposed. Consider first a two-class pattern recognition problem. We use one neuron (possibly higher order) with a sigmoid in the range from −1 to 1. Positive output means class 1 and negative output means class 2. The main idea of the method is that it iterates the weights in such a way as to move the decision boundary to a place of low pattern density. Constraining the length of the weight vector, if the neuron output is mostly near 1 or −1, then this means that the patterns are mostly far away from the decision boundary and we have probably a good classifier. We define a function which measures how close the output is to 1 or −1. Training is performed by a steepest-ascent algorithm on the weights. The method is extended to the multiclass case by applying the previous procedure in a hierarchical manner (i.e., by partitioning the patterns into two groups, then considering each group separately and partitioning it further and so on until we end up with the final classifier).

On a theoretical property of the bhattacharyya coefficient as a feature evaluation criterion

In the context evaluation of features in a two-class pattern recognition problem it is shown that, irrespective of the values of the a priori probabilities of the two classes, the maximum differences between the existing lower and upper bounds to Bayesian probability of error in terms of the Bhattacharyya coefficient is approximately equal to 0.2071.

An Optimal Set of Discriminant Vectors

A new method for the extraction of features in a two-class pattern recognition problem is derived. The main advantage is that the method for selecting features is based entirely upon discrimination or separability as opposed to the more common approach of fitting. The classical example of fitting is the use of the eigenvectors of the lumped covariance matrix corresponding to the largest eigenvalues. In an analogous manner, the new technique selects discriminant vectors (or features) corresponding to the largest discrim-values. The new method is compared to some of the more popular alternative techniques via both data-dependent and mathematical examples. In addition, a recursive method for obtaining the discriminant vectors is given.

The Reliability of Linear Feature Extractors

This paper introduces the concepts of capacity and reliability of a linear feature extractor. The relationship between the capacity introduced here and the channel capacity in information theory is discussed in some detail. The reliability is associated with the least favorable distributions and gives us a measure of the effectiveness for the worst possible case. It is of particular importance in pattern recognition problems because we have no control over the distributions of the patterns. It is shown that for the family of probability distributions with covariance matrix S, the most reliable (i.e., minimax) feature extractor is the Karhunan-Loève expansion. The concept of reliability is extended to the two-class pattern recognition problem and is discussed in terms of the Bhattacharyya distance.

Selection of Pattern Features by Mathematical Programming Algorithms

The selection of a "best" subset of features from a given set is considered for the case where a multiplicity of constraints are present in the two-class pattern recognition problem. The problem can be formulated with a Lagrangian multiplier and an efficient iterative technique used in conjunction with dynamic programming procedures to obtain a solution when two constraints are present. The problem can be formulated in integer programming terms and use made of available integer programming computer algorithms for solution when multiple constraints are present. The Lemke-Spielberg zero-one integer algorithm is particularly useful, but other integer programming algorithms can be used also.