Mayer Aladjem

Fingerprint image enhancement using filtering techniques

Extracting minutiae from fingerprint images is one of the most important steps in automatic fingerprint identification and classification. Minutiae are local discontinuities in the fingerprint pattern, mainly terminations and bifurcations. In this work we propose two methods for fingerprint image enhancement. The first one is carried out using local histogram equalization, Wiener filtering, and image binarization. The second method use a unique anisotropic filter for direct grayscale enhancement. The results achieved are compared with those obtained through some other methods. Both methods show some improvement in the minutiae detection process in terms of either efficiency or time required.

A Growing and Pruning Method for Radial Basis Function Networks

A recently published generalized growing and pruning (GGAP) training algorithm for radial basis function (RBF) neural networks is studied and modified. GGAP is a resource-allocating network (RAN) algorithm, which means that a created network unit that consistently makes little contribution to the networks performance can be removed during the training. GGAP states a formula for computing the significance of the network units, which requires a d-fold numerical integration for arbitrary probability density function p(x) of the input data x (x isin R d) . In this work, the GGAP formula is approximated using a Gaussian mixture model (GMM) for p(x) and an analytical solution of the approximated unit significance is derived. This makes it possible to employ the modified GGAP for input data having complex and high-dimensional p(x), which was not possible in the original GGAP. The results of an extensive experimental study show that the modified algorithm outperforms the original GGAP achieving both a lower prediction error and reduced complexity of the trained network.

On pattern classification with Sammon's nonlinear mapping an experimental study

Abstract Sammons mapping is conventionally used for exploratory data projection, and as such is usually inapplicable for classification. In this paper we apply a neural network (NN) implementation of Sammons mapping to classification by extracting an arbitrary number of projections. The projection map and classification accuracy of the mapping are compared with those of the auto-associative NN (AANN), multilayer perceptron (MLP) and principal component (PC) feature extractor for chromosome data. We demonstrate that chromosome classification based on Sammons (unsupervised) mapping is superior to the classification based on the AANN and PC feature extractor and highly comparable with that based on the (supervised) MLP. c 1998 Pattern Recognition Society.

Feature extraction by neural network nonlinear mapping for pattern classification

Feature extraction for exploratory data projection aims for data visualization by a projection of a high-dimensional space onto two or three-dimensional space, while feature extraction for classification generally requires more than two or three features. We study extraction of more than three features, using neural network (NN) implementation of Sammons mapping to be applied for classification. The experiments reveal that Sammons mapping, the multilayer perceptron (MLP) and the principal component analysis (PCA) based feature extractors yield similar classification performance. We investigate a random- and PCA-based initializations of Sammons mapping. When the PCA is applied to initialize Sammons projection, only one experiment is required and only a fraction of the training period is needed to achieve performance comparable with that of the random initialization. Furthermore, the PCA based initialization affords better human chromosome classification performance even when using a few eigenvectors.

A matching pursuit method for approximating overlapping ultrasonic echoes

Ultrasonic pulse-echo methods have been used extensively in measuring the thickness of layered structures as well as those of thin adhesive interface layers. When acoustically measuring thin layers, the resulting echoes from two successive interfaces overlap in time, limiting the minimum thickness that can be resolved using conventional pulse-echo techniques. In this paper, we propose a method, named support matching pursuit (SMP), for resolving the individual echoes. The method is based on the concept of sparse signal approximation in an overcomplete dictionary composed of Gabor atoms (elementary functions). Although the dictionary enables highly flexible approximations, it is also overcomplete, which implies that the approximation is not unique. We propose a method for approximation in which each ultrasonic echo is principally represented by a single atom and therefore has a physical interpretation. SMP operates similarly to the sparse matching pursuit (MP) method. It iteratively improves the approximation by adding, at each iteration, a single atom to the solution set. However, our atom selection criterion utilizes the time localization nature of ultrasonic echoes, which causes portions of a multi-echo ultrasonic signal to be composed mainly from a single echo. This leads to accurate approximations in which each echo is characterized by a set of physical parameters that represent the composing ultrasonic echoes. In the current research we compare SMP to other sparse approximation methods such as MP and basis pursuit (BP). We perform simulations and experiments on adhesively bonded structures which clearly demonstrate the superior performance of the SMP method over the MP and BP methods.

Multiclass discriminant mappings

Abstract Multiclass mappings oriented to the classification of multivariate data are discussed. Parametric and nonparametric mapping criteria are used for single level (one shot) projections. A method for reducing the computation complexity of the optimization procedure is suggested. In the case of a large number of classes, a sequential mapping based on the binary tree scheme is proposed. An algorithm for interactive binary tree design is presented. It combines automated and manual composition of the groups of the classes at the nodes of the tree. An example of the application of the tree mapping is presented.

Regularized discriminant analysis for face recognition

This paper studies regularized discriminant analysis (RDA) in the context of face recognition. We check RDA sensitivity to different photometric preprocessing methods and compare its performance to other classifiers. Our study shows that RDA is better able to extract the relevant discriminatory information from training data than the other classifiers tested, thus obtaining a lower error rate. Moreover, RDA is robust under various lighting conditions while the other classifiers perform badly when no photometric method is applied.

Nonparametric discriminant analysis via recursive optimization of Patrick-Fisher distance

A method for the linear discrimination of two classes is presented. It searches for the discriminant direction which maximizes the Patrick-Fisher (PF) distance between the projected class-conditional densities. It is a nonparametric method, in the sense that the densities are estimated from the data. Since the PF distance is a highly nonlinear function, we propose a recursive optimization procedure for searching the directions corresponding to several large local maxima of the PF distance. Its novelty lies in the transformation of the data along a found direction into data with deflated maxima of the PF distance and iteration to obtain the next direction. A simulation study and a medical data analysis indicate the potential of the method to find the sequence of directions with significant class separations.

Parametric and nonparametric linear mappings of multidimensional data

Abstract A new criterion for linear mapping of the samples from two classes is presented. Some parametric and nonparametric forms of the criterion are suggested. Based on them most of the known linear mapping projections can be created. New projections can be obtained as well. An experimental study with synthetic and real data is discussed. It confirms the effectiveness of the new mapping projections for the data with complicated classification structure.

Projection pursuit mixture density estimation

In this paper we seek a Gaussian mixture model (GMM) of an n-variate probability density function. Usually the parameters of GMMs are determined in the original n-dimensional space by optimizing a maximum likelihood (ML) criterion. A practical deficiency of this method of fitting GMMs is its poor performance when dealing with high-dimensional data since a large sample size is needed to match the accuracy that is possible in low dimensions. We propose a method for fitting the GMM based on the projection pursuit strategy. This GMM is highly constrained and hence its ability to model structure in subspaces is enhanced, compared to a direct ML fitting of a GMM in high dimensions. Our method is closely related to recently developed independent factor analysis (IFA) mixture models. The comparisons with ML fitting of GMM in n-dimensions and IFA mixtures show that the proposed method is an attractive choice for fitting GMMs using small sizes of training sets.