Helton Hideraldo Bíscaro

Hand movement recognition for Brazilian Sign Language: A study using distance-based neural networks

In this paper, the vision-based hand movement recognition problem is formulated for the universe of discourse of the Brazilian Sign Language. In order to analyze this specific domain we have used the artificial neural networks models based on distance, including neural-fuzzy models. The experiments explored here show the usefulness of these models to extract helpful knowledge about the classes of movements and to support the project of adaptative recognizer modules for Libras-oriented computational tools. Using artificial neural networks architectures - Self Organizing Maps and (Fuzzy) Learning Vector Quantization, it was possible to understand the data space and to build models able to recognize hand movements performed for one or more than one specific Libras users.

Simulation of soft tissue deformation: A new approach

An approach is presented in this paper, combining methods and models that are efficient enough to simulate elastic deformation, obtaining equilibrium between visual and haptic realism. Many medical training computational applications manipulate 3D objects that represent organs and human tissues. These representations, in function of the training requirements for which they are meant, may include parameter such as shape, topology, color, volume texture and, in certain cases, physical properties such as elasticity and stiffness. Based on this model, visual and/or haptic outputs are generated for users, and need to be realistic. In other words, they need to provide the learner with sensations close enough to those they would have if the training were provided with real life objects. However, its computational cost is too high to simultaneously provide visual and haptic realism in real-time. The results from deformation response time are compatible with those required for haptic interaction and the visual results from using meshes composed of a large number of polygons.

Using Bipartite Graphs for 3D Cardiac Model Retrieval

Three-dimensional models have been used to aid medical diagnoses, using images generated by modalities like Magnetic Resonance Imaging. They can provide a more complete vision of objects since their depth is taken into account. Content-based Image Retrieval (CBIR) has also been used to aid the diagnosis. One important step in Three-dimensional CBIR (Model Retrieva) systems is the comparison between two models by using a set of features extracted and stored in a database. In this paper we present a novel method to compare two models, using the Bipartite graphs technique, with the aim to improve the retrieval precision. This technique retrieves 3D medical models of the left ventricle in order to aid the diagnosis of Congestive Heart Failure. Results showed that the novel method improved the precision by 10% when compared to the Similarity Function of Euclidean and Manhattan distance. These results confirmed that bipartite graph techniques can be used to improve the accuracy of Model Retrieval systems.

Realistic Simulation of Deformation for Medical Training Applications

Many computational applications for virtual medical training manipulate 3D objects that represent organs and human tissues. These representations, in function of the training requirements, may include parameters such as shape, topology, color, volume, texture and, in certain cases, physical properties such as elasticity and stiffness. An approach is presented in this paper, combining methods and models that are efficient to simulate elastic deformation, obtaining equilibrium between visual and haptic realism. Based on this model, visual and/or haptic outputs are generated for users in order to provide to learner sensations close enough to those they would have if the training were provided with real objects. The results regarding deformation processing time are compatible with those required for haptic interaction and provide adequate visual feedback from using meshes composed of a large number of polygons.

Tutorial sobre Fuzzy-c-Means e Fuzzy Learning Vector Quantization: Abordagens Híbridas para Tarefas de Agrupamento e Classificação

Neste tutorial e apresentada uma discussao sobre o algoritmo Fuzzy-c-Means e sobre as Redes Neurais Fuzzy, considerando a proposta de insercao de principios da Teoria de Conjuntos Fuzzynas abordagens de agrupamento e classificacao classicas: algoritmo c-Means e o modelo neural Learning Vector Quantization. A motivacao para a construcao de um modelo hibrido, dessa categoria, e conferir as abordagens classicas a capacidade de lidar adequadamente com aspectos de incerteza e imprecisao, comumente encontrados em problemas reais.

A New Descriptor for Retrieving 3D Objects Applied in Congestive Heart Failure Diagnosis

Content-Based Image Retrieval (CBIR) aims to retrieve similar graphical objects from large databases based on their contents. CBIR requires definition of descriptors, algorithms that condense information from the object in order to represent it usually as a real number or a vector in Rn. This article presents the Spectral Descriptor, a new descriptor designed for retrieving three-dimensional geometric objects applied to aid the diagnosis of Congestive Heart Failure (CHF). Our descriptor is based on techniques of compressive sensing and rewrites the coordinates of 3D objects vertices on a basis on which they have a sparse representation. Tests with surfaces reconstructed from heart MRI images, specifically from left ventricle, show that the descriptor has presented a good performance, reaching an average precision of approximately 85% for CHF and 71% for non-CHF cases, maintaining high levels of precision. Results also showed that the Spectral Descriptor can decrease the high dimensionality of features vectors in CBIR systems.

Camera calibration for sport images: Using a modified RANSAC-based strategy and genetic algorithms

Camera calibration is an attractive problem that have received relatively great attention from scientists in the last years. It has several interesting applications, particularly in sports broadcasting technology. We present in this paper a complete system that receives a sport image, automatically detects control points in the image with a RANSAC-based strategy and solves the optimization problem by using a genetic algorithm. In fact, our approach introduces an improvement in the RANSAC-based strategy which is able to refine its results, and shows that a simple genetic algorithm is capable to solve the camera calibration problem with relatively small computational effort.

TETRAHEDRON TOPOLOGICAL CHARACTERIZATION WITH APPLICATION IN VOLUMETRIC RECONSTRUCTION

One of the challenges of Volume Modelling is the definition of theoretical frameworks to support object manipulation and representation. Opposite to what is seen in the field of surface modelling, few satisfactory tools have been presented that ensure topological robustness to volumetric models. In this paper we offer one such mathematical framework for volumetric model definition based on tetrahedral meshes. From a complete tetrahedron topological characterization, a set of Morse Operators is given, which enable global topological control during tetrahedron addition or removal. A number of applications can be envisaged, from geometrical modelling to volume reconstruction. We show the effectiveness of the tetrahedron characterization framework for volume reconstruction from images, showing that the method is capable of handling certain types of noise topologically without the need for a time consuming preprocessing step, or a post-processing step to detect cavities and holes.