Evandro Ottoni Teatini Salles

Artificial neural network for identification of a substance from a Mössbauer data bank

Abstract Mossbauer data and references of the minerals reported in the literature have been stored in a computer. Artificial neutral networks (ANN) were taught with the average values of experimental data of isomer shift quadrupole splitting of known mineral systems (sulphate, sulphide and sulphites, and silicates). Artificial neural networks successfully identified the unknown substance when fed with the new values of isomer shift and quadrupole splitting.

Super-Resolution Image Reconstruction Using Nonparametric Bayesian INLA Approximation

Super-resolution (SR) is a technique to enhance the resolution of an image without changing the camera resolution, through using software algorithms. In this context, this paper proposes a fully automatic SR algorithm, using a recent nonparametric Bayesian inference method based on numerical integration, known in the statistical literature as integrated nested Laplace approximation (INLA). By applying such inference method to the SR problem, this paper shows that all the equations needed to implement this technique can be written in closed form. Moreover, the results of several simulations (three of them are here presented) show that the proposed algorithm performs better than other SR algorithms recently proposed. As far as the authors know, this is the first time that the INLA is used in the area of image processing, which is a meaningful contribution of this paper.

Topographic independent component analysis based on fractal theory and morphology applied to texture segmentation

Texture analysis and segmentation is an important area in image processing. One can employ texture segmentation for quality control in processes related to skin-leather, textile or marble/granite industries, for example. In such a context, the topographic independent component analysis (TICA) is presented as a technique for texture segmentation in which the image base is obtained from the mixture matrix of the model by implementing a bank of statistical filters, which are capable to capture the inherent properties of each texture. Indeed, using the energy as topographic criterion, the TICA filter bank exhibits results that are similar to the independent component analysis (ICA) model, as it has been already shown in the literature. In this paper, we show that using energy and morphologic fractal texture descriptors as topographic criterion those results are improved, in the sense that the segmentation error and the amount of filters are reduced, for the same textures.

Human automatic detection and tracking for outdoor video

Human automatic tracking is still a problem. This paper presents an approach to treat this problem, where neither the target color model nor the background color model are initially supposed to be known. Besides, the illumination conditions and background may change and, furthermore, the target can be occluded for a determined time. In such scenario, this work proposes an approach for human tracking in outdoor environment by identifying the target. A modified version of Mean Shift is employed, in which it is used HSV color space and a procedure to update its color model. In addition, a method to find the face of the person in video sequence is also presented, in order to identify people or to validate that the target is a human. For evaluation, four metrics were proposed and it has been carried out a series of experiments, and also the presented method was superior to other techniques evaluated.

An Evolving System Based on Probabilistic Neural Network

Although traditional techniques of machine learning have, in many cases, presented good results, they have been inefficient for data which are constantly expanding and changing over time. To address these problems, new learning techniques have been proposed in the literature. In this paper we propose a technique called ePNN presenting aspects of this recent paradigm of learning. We carried out a series of experiments that showed its efficiency over previous approaches.

An incremental neural network with a reduced architecture

This paper proposes a technique, called Evolving Probabilistic Neural Network (ePNN), that presents many interesting features, including incremental learning, evolving architecture, the capacity to learn continually throughout its existence and requiring that each training sample be used only once in the training phase without reprocessing. A series of experiments was performed on data sets in the public domain; the results indicate that ePNN is superior or equal to the other incremental neural networks evaluated in this paper. These results also demonstrate the advantage of the small ePNN architecture and show that its architecture is more stable than the other incremental neural networks evaluated. ePNN thus appears to be a promising alternative for a quick learning system and a fast classifier with a low computational cost.

Automatic sleep staging using a single-channel EEG modeling by Kalman Filter and HMM

This paper proposes an automatic classification system for sleep stage of persons. The sleep condition of a person is monitored by one channel electroencephalogram (EEG). Because of the non stationary nature of the signal, for the feature extraction task, it is used the coefficients of a Kalman Filter modeling. The classification task is realized by a K-Means Segmental HMM (Hidden Markov Model). To evaluate the performance of the system, it is used the MIT-BIH Polysomnographic EEG database. At the end, the results are presented and discussed.

Unsupervised color image segmentation based on local fractal descriptor and J-images

This paper proposes an improved version for the JSEG color image segmentation algorithm, combining the classical JSEG algorithm and a local fractal operator that measures the fractal dimension of each pixel, thus improving the boundary detection in the J-map. Experiments with natural color images of the Berkeley Segmentation Dataset and Benchmark are presented, which show improved results in comparison with the classical JSEG algorithm.

Fractal-JSEG: JSEG Using an Homogeneity Measurement Based on Local Fractal Descriptor

This paper proposes an improved version for the JSEG color image segmentation algorithm, combining the classical JSEG algorithm and a local fractal operator that measures the fractal dimension of each pixel, thus improving the boundary detection in the J-map. Experiments with natural color images of the Berkeley Segmentation Dataset and Benchmark are presented, which show improved results in comparison with the classical JSEG algorithm.

Multi-label incremental learning applied to web page categorization

Multi-label problems are challenging because each instance may be associated with an unknown number of categories, and the relationship among the categories is not always known. A large amount of data is necessary to infer the required information regarding the categories, but these data are normally available only in small batches and distributed over a period of time. In this work, multi-label problems are tackled using an incremental neural network known as the evolving Probabilistic Neural Network (ePNN). This neural network is capable of continuous learning while maintaining a reduced architecture, so that it can always receive training data when available with no drastic growth of its structure. We carried out a series of experiments on web page data sets and compared the performance of ePNN to that of other multi-label categorizers. On average, ePNN outperformed the other categorizers in four out of five metrics used for evaluation, and the structure of ePNN was less complex than that of the other algorithms evaluated.