Catálogo de publicaciones - libros

In this work we tackle the task of detecting biological motifs, i.e. subsequences with an associated function. This task is important in bioinformatics because it is related to the prediction of the behaviour of the whole protein. Artificial neural networks are used to, somewhat, translate the sequence of amino acids of the protein into a code that shows the subsequences where the presence of the studied motif is expected. The experimentation performed prove the good performance of our approach.

This paper reports the results obtained in the implementation of an Optical Braille Recognizer (O.B.R.), as well as the construction of a keyboard for the Braille code. This project was developed with the objective of enabling teachers of blind people, who do not know the Braille code, to visualize the texts written by their students. An electronic keyboard, less noisy and less expensive than the traditional mechanical ones was built too. To achieve these objectives, the ”Compendium of the Braille code for the Portuguese Language” was used. The final program translates plain text, mathematics and chemistry sheets written in Braille code. It’s also possible to write plain text, mathematics or chemistry using the developed keyboard. The program is written in Java and the keyboard communicates with it through serial port.

In this paper is presented a method for the automatic DNA spots classification and extraction of profiles associated in DNA polyacrilamide gel electrophoresis based on image processing. A software which implements this method was developed, composed by four modules: Digital image acquisition, image preprocessing, feature extraction and classification, and DNA profile extraction. The use of different types of algorithms as: C4.5 Decision Trees, Support Vector Machines and Leader Algorithm are needed to resolve all the tasks. The experimental results show that this method has a very nice computational behavior and effectiveness, and provide a very useful tool to decrease the time and increase the quality of the specialist responses.

In the paper we propose a method based on Bayesian framework for selecting the best kernel function for supervised learning problem. The parameters of the kernel function are considered as model parameters and maximum evidence principle is applied for model selection. We describe a general scheme of Bayesian regularization, present model of kernel classifiers as well as our approximations for evidence estimation, and then give some results of experimental evaluation.

One of the basic problems of applied mathematics is to find a synthetic expression (model) which captures the essence of a system given a (necessarily) finite sample which reflects selected characteristics. When the model considers several independent variables its mathematical treatment may become burdensome or even downright impossible from a practical standpoint.

We describe the application of 1-Dimensional Kohonen Networks in the classification of color 2D images which has been evaluated in Popocatépetl Volcano’s images. The Popocatépetl, located in the limits of the State of Puebla in México, is active and under monitoring since 1997. We will consider one of the problems related with the question if our application of the Kohonen Network classifies according to the total intensity color of an image or well, if it classifies according to the connectivity, i.e. the topology, between the pixels that compose an image. In order to give arguments that support our hypothesis that our procedures share the classification according to the topology of the pixels in the images, we will present two approaches based a) in the evaluation of the classification given by the network when the pixels in the images are permuted; and,b) when an additional metric to the Euclidean distance is introduced.

A data dependent wavelet transform based on the modified lifting scheme is presented. The algorithm is based on the wavelet filters derived from a generalized lifting scheme. The proposed framework for the lifting scheme permits to obtain easily different wavelet FIR filter coefficients in the case of the (~N, N) lifting. To improve the performance of the lifting filters the presented technique additionally realizes IIR filtering by means of the feedback to the already calculated wavelet coefficients. The perfect image restoration in this case is obtained employing the particular features of the lifting scheme. Changing wavelet FIR filter order and/or FIR and IIR coefficients, one can obtain the filter frequency response that match better to the image data than the standard lifting filters, resulting in higher data compression rate. The designed algorithm was tested on different images. The obtained simulation results show that the proposed method performs better in data compression for various images in comparison to the standard technique resulting in significant savings in compressed data length.

A new robust matching algorithm for motion detection and computation of precise estimates of motion vectors of moving objects in a sequence of images is presented. Common matching algorithms of dynamic image analysis usually utilize local smoothness constraints. The proposed method exploits global motion smoothness. The suggested matching algorithm is robust to motion discontinuity as well as to noise degradation of a signal. Computer simulation and experimental results demonstrate an excellent performance of the method in terms of dynamic motion analysis.

In this paper we introduce a general framework for hierarchical clustering that deals with both static and dynamic data sets. From this framework, different hierarchical agglomerative algorithms can be obtained, by specifying an inter-cluster similarity measure, a subgraph of the -similarity graph, and a cover algorithm. A new clustering algorithm called and its dynamic version are presented, which are specific versions of the proposed framework. Our evaluation experiments on several standard document collections show that this algorithm requires less computational time than standard methods in dynamic data sets while achieving a comparable or even better clustering quality. Therefore, we advocate its use for tasks that require dynamic clustering, such as information organization, creation of document taxonomies and hierarchical topic detection.

The automatic detection of different foot’s diseases requires the analysis of a footprint, obtained from a digital image of the sole. This paper shows that optical monochromatic images are not suitable for footprint segmentation purposes, while color images provide enough information for carrying out an efficient segmentation. It is shown that a multiplayer perceptron trained with bayesian regularization backpropagation allows to adequately classify the pixels on the color image of the footprint and in this way, to segment the footprint without fingers. The footprint is improved by using a classical smoothing filter, and segmented by performing erosion and dilation operations. This result is very important for the development of a low cost system designed to diagnose pathologies related to the footprint form.