Mauricio Kugler

Feature subset selection for support vector machines using confident margin

The aim of this study is to develop a feature subset selection (FSS) method based on the margin of support vector machines (SVM). The problem of directly using the SVM margin is that it does not always provide clear relationship between its value and the performance of SVM, and the best obtained subset is not guaranteed to be the best possible one. In this paper, a new solution is describe by the introduction of the confident margin (CM) in the subset criterion, which permits to get near the best recognition rate by monitoring the peak of CM curve without directly calculating the recognition rate, in order to save computational time. The performance of the proposed method was evaluated in artificial and real-world data experiments.

A Sound Localization and Recognition System using Pulsed Neural Networks on FPGA

Pulsed neurons are suitable for processing time series data, like sound signals, and can be easily implemented in hardware. In this paper, we propose an aural information processing system based on the human auditory system using a pulsed neuron model and a correspondent implementation in an FPGA device. Experimental results show that an FPGA based implementation of the proposed system can successfully identify the results faster than a similar software implementation. Noise tolerance experimental results are also presented.

CombNET-III: A Support Vector Machine Based Large Scale Classifier with Probabilistic Framework

Several research fields have to deal with very large classification problems, e.g. handwritten character recognition and speech recognition. Many works have proposed methods to address problems with large number of samples, but few works have been done concerning problems with large numbers of classes. CombNET-II was one of the first methods proposed for such a kind of task. It consists of a sequential clustering VQ based gating network (stem network) and several Multilayer Perceptron (MLP) based expert classifiers (branch networks). With the objectives of increasing the classification accuracy and providing a more flexible model, this paper proposes a new model based on the CombNET-II structure, the CombNET-III. The new model, intended for, but not limited to, problems with large number of classes, replaces the branch networks MLP with multiclass Support Vector Machines (SVM). It also introduces a new probabilistic framework that outputs posterior class probabilities, enabling the model to be applied in different scenarios (e.g. together with Hidden Markov Models). These changes permit the use of a larger number of smaller clusters, which reduce the complexity of the final classifiers. Moreover, the use of binary SVM with probabilistic outputs and a probabilistic decoding scheme permit the use of a pairwise output encoding on the branch networks, which reduces the computational complexity of the training stage. The experimental results show that the proposed model outperforms both the previous model CombNET-II and a single multiclass SVM, while presenting considerably smaller complexity than the latter. It is also confirmed that CombNET-III classification accuracy scales better with the increasing number of clusters, in comparison with CombNET-II.

A Complete Hardware Implementation of an Integrated Sound Localization and Classification System Based on Spiking Neural Networks

Several applications would emerge from the development of artificial systems able to accurately localize and identify sound sources. This paper proposes an integrated sound localization and classification system based on the human auditory system and a respective compact hardware implementation. The proposed models are based on spiking neurons, which are suitable for processing time series data, like sound signals, and can be easily implemented in hardware. The system uses two microphones, extracting the time difference between the two channels with a chain of coincidence detection spiking neurons. A spiking neural networks process the time-delay pattern, giving a single directional output. Simultaneously, an independent spiking neural network process the spectral information of on audio channel in order to classify the source. Experimental results show that a the proposed system could successfully locate and identify several sound sources in real time with high accuracy.

A novel approach for hardware based sound localization

Sound localization is an important ability intrinsic to animals, being currently explored by several researches. Even though several systems and implementations have being proposed, the majority is very complex and not suitable for embedded systems. This paper proposes a new approach for binaural sound localization and the corresponding implementation in an Field Programable Gate Array (FPGA) device. The system is based on the signal processing modules of a previously proposed sound processing system, which converts the input signal to spike trains. The time difference extraction and feature generation methods introduced in this paper create simple binary feature vectors, used as training data for a standard LVQ neural network. An output temporal layer uses the time information of the sound signals in order to reduce the misclassifications of the classifier. Preliminary experimental results show high accuracy with small logic and memory requirements.

A Novel Approach for Hardware Based Sound Classification

Several applications would emerge from the development of efficient and robust sound classification systems able to identify the nature of non-speech sound sources. This paper proposes a novel approach that combines a simple feature generation procedure, a supervised learning process and fewer parameters in order to obtain an efficient sound classification system solution in hardware. The system is based on the signal processing modules of a previously proposed sound processing system, which convert the input signal in spike trains. The feature generation method creates simple binary features vectors, used as the training data of a standard LVQ neural network. An output temporal layer uses the time information of the sound signals in order to eliminate the misclassifications of the classifier. The result is a robust, hardware friendly model for sound classification, presenting high accuracy for the eight sound source signals used on the experiments, while requiring small FPGA logic and memory resources.

CombNET-III with Nonlinear Gating Network and Its Application in Large-Scale Classification Problems

Modern applications of pattern recognition generate very large amounts of data, which require large computational effort to process. However, the majority of the methods intended for large-scale problems aim to merely adapt standard classification methods without considering if those algorithms are appropriated for large-scale problems. CombNET-II was one of the first methods specifically proposed for such kind of a task. Recently, an extension of this model, named CombNET-III, was proposed. The main modifications over the previous model was the substitution of the expert networks by Support Vectors Machines (SVM) and the development of a general probabilistic framework. Although the previous models performance and flexibility were improved, the low accuracy of the gating network was still compromising CombNET-IIIs classification results. In addition, due to the use of SVM based experts, the computational complexity is higher than CombNET-II. This paper proposes a new two-layered gating network structure that reduces the compromise between number of clusters and accuracy, increasing the models performance with only a small complexity increase. This high-accuracy gating network also enables the removal the low confidence expert networks from the decoding procedure. This, in addition to a new faster strategy for calculating multiclass SVM outputs significantly reduced the computational complexity. Experimental results of problems with large number of categories show that the proposed model outperforms the original CombNET-III, while presenting a computational complexity more than one order of magnitude smaller. Moreover, when applied to a database with a large number of samples, it outperformed all compared methods, confirming the proposed models flexibility.

A new approach for applying support vector machines in multiclass problems using class groupings and truth tables

The Support Vector Machines (SVMs) had been showing a high capability of complex hyperplane representation and great generalization power. These characteristics lead to the development of more compact and less computational complex methods than the One-versus-Rest (OvR) and One-versus-One (OvO) [1] classical methods in the application of SVMs in multiclass problems. This paper proposes a new method for this task, named Truth Table Fitting Multiclass SVM (TTF-MCSVM), in which less SVMs are used than other classical methods. The main objective of this research is the development of an efficient method to be applied in problems with very large number of classes, like in the recognition of East Asian languages characters (e.g. Japanese and Chinese kanji).

A novel approach for sound approaching detection

The detection of approaching vehicles is a very important topic on the development of complementary traffic safety systems. However, the majority of the proposed approaches are very complex and not suitable for embedded applications. This paper proposes a new sound approaching detection algorithm specifically intended for hardware implementation. Experimental results show higher accuracy and earlier detection when comparing to other methods.

Accurate 3D Reconstruction of a Whole Pancreatic Cancer Tumor from Pathology Images with Different Stains

When applied to 3D image reconstruction, conventional landmark-based registration methods tend to generate unnatural vertical structures due to inconsistencies between the employed model and the real tissue. This paper demonstrates a fully non-rigid image registration method for 3D image reconstruction which considers the spatial continuity and smoothness of each constituent part of the microstructures in the tissue. Corresponding landmarks are detected along the images, defining a set of trajectories, which are smoothed out in order to define a diffeomorphic mapping. The resulting reconstructed 3D image preserves the original tissue architecture, allowing the observation of fine details and structures.