Made available in DSpace on 2016-08-17T14:53:22Z (GMT). No. of bitstreams: 1 dissertacao Luis Claudio.pdf: 2055680 bytes, checksum: 4e51c93fd4aa6edfad51ab54ad4044f6 (MD5) Previous issue date: 2012-07-24 / This thesis proposes the modeling and implementation of a telediagnostic system for analysis and detection of lesions in mammographic images based on independent component analysis and support vector machine. The system analyzes images from digital mammography sent over the Internet and provides a diagnostic, indicating the presence of suspicious regions, which can be confirmed by a specialist in mammographic images. Besides presenting the methodology for the development of the proposed system, a prototype was developed for testing and to measure its efficiency. The database used for training and testing of the algorithms is the mini-MIAS, and was employed independent component analysis to extract the filters used in segmenting the regions of interest, as well support vector machine to classify regions of interest in normal or suspicious. From tests with the database used, we obtained an average accuracy of 87.8% for images containing lesions. / Este trabalho propõe a modelagem e implementação de um sistema de telediagnóstico para análise e detecção automática de lesões em imagens mamográficas, baseado em análise de componentes independentes e máquina de vetor de suporte. O sistema analisa imagens de mamografia digital enviadas pela Internet e fornece um diagnóstico da imagem, indicando a presença de regiões suspeitas, que podem ser confirmadas por um especialista em imagens mamográficas. Além de apresentar a metodologia para o desenvolvimento do sistema proposto, foi desenvolvido um protótipo para a realização de testes objetivando medir sua eficiência. A base de dados usada para treinamento e teste dos algoritmos foi a mini-MIAS, e foi empregada análise de componentes independentes para extrair os filtros usados na segmentação das regiões de interesse, bem como máquina de vetor de suporte para classificar as regiões de interesse em normais ou suspeitas. A partir de testes realizados com a base de dados utilizada, obteve-se média de acerto de 87,8% para imagens que contém lesões.

Orientadores: Romis Ribeiro de Faissol Attux, Leonardo Tomazeli Duarte / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-23T17:30:21Z (GMT). No. of bitstreams: 1 Ando_RafaelAssato_M.pdf: 1700685 bytes, checksum: 9befef5632e55e27a24af3e73c648c47 (MD5) Previous issue date: 2013 / Resumo: Neste trabalho, estudamos o problema de BSS no contexto de misturas polinomiais sob três perspectivas: uma teórica - voltada ao estudo de separabilidade estrutural -, uma ligada à proposta de novas abordagens - especialmente como extensões de metodologias baseadas em redes recorrentes - e uma relacionada ao tratamento de problemas práticos como redução do efeito show-through na digitalização de documentos. A primeira dessas perspectivas levou à proposta de uma nova abordagem do problema de separação não-linear baseada numa formulação do problema instantâneo de inversão como uma tarefa de solução de um sistema de equações algébricas não-lineares. Essa abordagem levou à proposição de novos métodos para lidar com o problema LQ e também pode ser aplicada a outros modelos de mistura. A segunda perspectiva levou à construção de um arcabouço para tratamento do problema LQ baseado numa rede imunológica artificial, o qual trouxe uma menor demanda por informação a priori sobre o problema e provê maior robustez em termos de convergência global. Por fim, a aplicação do ferramental desenvolvido a problemas práticos de tratamento de imagens levou a um desempenho bastante satisfatório, encorajando a extensão futura para outros cenários de teste (como sensores químicos) / Abstract: In this work, the BSS problem in the context of polynomial mixtures will be studied under three perspectives: a theoretical one, regarding the structural separability analysis; another related to the proposal of new methodologies - especially as extensions of algorithms based on recurrent networks - and finally, one regarding the solutions to real world problems, such as the reduction of the show-through effect produced by digitally scanning documents. The first such perspectives led to the proposal of a new approach to the nonlinear BSS problem, based on a formulation to the instantaneous inversion problem as the solution of a non-linear algebraic equation system. This approach led to the proposal of new methods to deal with the LQ problem, which may also be applied to other mixing models. The second perspective led to the development of an algorithm based on artificial immune system (AIS) to solve the LQ model, requiring less a priori information about the problem and providing better robustness in terms of global convergence. Finally, the application of the pro-posed methods to the practical problem of image treatment presented a very satisfactory performance, encouraging the possible extension to other test scenarios in the future, such as chemical sensors / Mestrado / Engenharia de Computação / Mestre em Engenharia Elétrica

Orientadores: Romis Ribeiro de Faissol Attux, Jugurta Rosa Montalvão Filho / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-23T23:31:44Z (GMT). No. of bitstreams: 1 Silva_DanielGuerreiroe_D.pdf: 5960509 bytes, checksum: febb8228109537e82dfcce66fca8aae8 (MD5) Previous issue date: 2013 / Resumo: Esta tese de doutorado possui como tema geral o desenvolvimento de algoritmos de Aprendizado de Máquina Baseado na Teoria da Informação (ITL - Information Theoretic Learning). O paradigma de ITL propõe o uso de critérios de treinamento baseados em medidas como entropia e informação mútua, em substituição aos tradicionais critérios baseados em estatísticas de segunda ordem. Os problemas de inversão cega de sistemas de Wiener e separação cega de sinais em corpos de Galois são os objetos de estudo e desenvolvimento dessas ferramentas. Estes problemas apresentam características marcantes quanto à necessidade de descritores estatísticos de ordem superior, por isso, apresenta-se uma série de contribuições que se baseiam em critérios de ITL e empregam algoritmos imuno-inspirados (ou heurísticas de busca) para adaptar os parâmetros dos modelos envolvidos. As propostas desenvolvidas abrem a perspectiva de futuras aplicações em áreas como genômica, codificação e geofísica. Espera-se também que os resultados apresentados fortaleçam um entendimento mais amplo de ITL, a fim de abranger estratégias mais genéricas de busca, de estimação de informação e de modelagem de dados / Abstract: The main theme of this thesis is the development of Information Theoretic Learning (ITL) algorithms. The ITL paradigm proposes the adoption of training criteria based on information measures such as entropy and mutual information, instead of traditional criteria based on second order statistics. The problems of blind inversion of Wiener systems and blind separation of signals over Galois fields are the tasks over which these tools are applied. Such problems present key aspects that establish a demand for higher order statistics, hence we present several contributions that are based on ITL criteria and employ immune-inspired algorithms (or heuristic-based methods) to perform the adaptation of the parameters of each related model. The developed proposals open the perspective of future applications in genomic, coding theory and geophysics. Furthermore, we expect that the presented results support a wider understanding of ITL, in order to comprise more general strategies of search, information estimation and data modeling / Doutorado / Engenharia de Computação / Doutor em Engenharia Elétrica

La thèse s’inscrit dans un contexte d’essor de l’audio spatialisé (5.1, Dolby Atmos...). Parmi les formats audio 3D existants, l’ambisonie permet une représentation spatiale homogène du champ sonore et se prête naturellement à des manipulations : rotations, distorsion du champ sonore. L’objectif de cette thèse est de fournir un outil d’analyse et de manipulation de contenus audio (essentiellement vocaux) au format ambisonique. Un fonctionnement temps-réel et en conditions acoustiques réelles sont les principales contraintes à respecter. L’algorithme mis au point est basé sur une analyse en composantes indépendantes (ACI) appliquée trame à trame qui permet de décomposer le champ acoustique en un ensemble de contributions, correspondant à des sources (champ direct) ou à de la réverbération. Une étape de classification bayésienne, appliquée aux composantes extraites, permet alors l’identification et le dénombrement des sources sonores contenues dans le mélange. Les sources identifiées sont localisées grâce à la matrice de mélange obtenue par ACI, pour fournir une cartographie de la scène sonore. Une étude exhaustive des performances est menée sur des contenus réels en fonction de plusieurs paramètres : nombre de sources, environnement acoustique, longueur des trames, ou ordre ambisonique utilisé. Des résultats fiables en terme de localisation et de comptage de sources ont été obtenus pour des trames de quelques centaines de ms. L’algorithme, exploité comme prétraitement dans un prototype d’assistant vocal domestique, permet d’améliorer significativement les performances de reconnaissance, notamment en prise de son lointaine et en présence de sources interférentes. / The context of this thesis is the development of spatialized audio (5.1 contents, Dolby Atmos...) and particularly of 3D audio. Among the existing 3D audio formats, Ambisonics and Higher Order Ambisonics (HOA) allow a homogeneous spatial representation of a sound field and allows basics manipulations, like rotations or distorsions. The aim of the thesis is to provides efficient tools for ambisonics and HOA sound scene analyse and manipulations. A real-time implementation and robustness to reverberation are the main constraints to deal with. The implemented algorithm is based on a frame-by-frame Independent Component Analysis (ICA), wich decomposes the sound field into a set of acoustic contributions. Then a bayesian classification step is applied to the extracted components to identify the real sources and the residual reverberation. Direction of arrival of the sources are extracted from the mixing matrix estimated by ICA, according to the ambisonic formalism, and a real-time cartography of the sound scene is obtained. Performances have been evaluated in different acoustic environnements to assess the influence of several parameters such as the ambisonic order, the frame length or the number of sources. Accurate results in terms of source localization and source counting have been obtained for frame lengths of a few hundred milliseconds. The algorithm is exploited as a pre-processing step for a speech recognition prototype and allows a significant increasing of the recognition results, in far field conditions and in the presence of noise and interferent sources.

Les aires cérébrales entretiennent des relations fonctionnelles, formant ainsi des réseaux qui peuvent être altérés dans diverses pathologies. L'étude de ces réseaux de connectivité fonctionnelle pourrait potentiellement aider au diagnostic d'un individu et au traitement de sa pathologie. À travers quatre études, nous avons montré que l'analyse en composantes indépendantes spatiale est une méthode suffisamment sensible, reproductible et spécifique pour mettre en évidence, à l’échelle individuelle et au repos, des réseaux sains et pathologiques fiables. Ainsi, le réseau pathologique sous-tendant les hallucinations acoustico-verbales permet de définir les aires cérébrales à traiter par stimulation magnétique transcrânienne. Parmi les réseaux sains, ceux qui impliquent le cortex cingulaire postérieur et le précunéus semblent profondément altérés dans les troubles de la conscience, et peuvent servir d'outil diagnostic pour distinguer le locked-in syndrome de l'état végétatif. Il est désormais possible de cartographier, à l'échelle individuelle, les relations fonctionnelles entre les aires cérébrales. L’étude à venir de la dynamique et du niveau d’activité des réseaux de connectivité fonctionnelle nous renseignera davantage sur leurs fonctions et leur implication dans la pathologie. / Brain areas are functionally connected in networks, even at rest. Since such connectivity networks could be impaired in several pathologies, they could potentially serve for diagnosis and treatment. Based on four studies, spatial independent component analysis has shown sufficient sensitivity, reproducibility and specificity to produce reliable healthy as well as pathological networks at the individual level. Therefore, the network underlying auditory hallucination could define the brain areas to treat by transcranial magnetic stimulation. Among the common resting-state networks, the ones involving the posterior cingular cortex and the precuneus seem deeply altered in disorders of consciousness, and so could be used as a diagnostic tool to distinguish the locked-in syndrome from the vegetative state. We can now map, at the individual level, the functional relationships between brain areas. Further studies on the dynamic and on the level of activity of the functional connectivity networks might provide relevant information about their functions and their involvement in the pathology.

This dissertation introduces a new approach for assessing the effects of pediatric epilepsy on the language connectome. Two novel data-driven network construction approaches are presented. These methods rely on connecting different brain regions using either extent or intensity of language related activations as identified by independent component analysis of fMRI data. An auditory description decision task (ADDT) paradigm was used to activate the language network for 29 patients and 30 controls recruited from three major pediatric hospitals. Empirical evaluations illustrated that pediatric epilepsy can cause, or is associated with, a network efficiency reduction. Patients showed a propensity to inefficiently employ the whole brain network to perform the ADDT language task; on the contrary, controls seemed to efficiently use smaller segregated network components to achieve the same task. To explain the causes of the decreased efficiency, graph theoretical analysis was carried out. The analysis revealed no substantial global network feature differences between the patient and control groups. It also showed that for both subject groups the language network exhibited small-world characteristics; however, the patient’s extent of activation network showed a tendency towards more random networks. It was also shown that the intensity of activation network displayed ipsilateral hub reorganization on the local level. The left hemispheric hubs displayed greater centrality values for patients, whereas the right hemispheric hubs displayed greater centrality values for controls. This hub hemispheric disparity was not correlated with a right atypical language laterality found in six patients. Finally it was shown that a multi-level unsupervised clustering scheme based on self-organizing maps, a type of artificial neural network, and k-means was able to fairly and blindly separate the subjects into their respective patient or control groups. The clustering was initiated using the local nodal centrality measurements only. Compared to the extent of activation network, the intensity of activation network clustering demonstrated better precision. This outcome supports the assertion that the local centrality differences presented by the intensity of activation network can be associated with focal epilepsy.

The objective of the present thesis research is to apply state-of-the-art experimental and data analysis techniques to the study of gas-liquid pipe flows, with a focus on conditions occurring in header-feeder systems of nuclear reactors under different accident scenarios. Novel experimental techniques have been proposed for the identification of the flow regime and measurement of the flow rates of both phases in gas-liquid flows. These techniques were automated, non-intrusive and economical, which ensured that their use would be feasible in industrial as well as laboratory settings. Measurements of differential pressure and the gas and liquid flow rates were collected in vertical upwards air-water flow at near-atmospheric pressure. It was demonstrated that the probability density function of the normalized differential pressure was indicative of the flow regime and using non-linear dimensionality reduction (the Elastic Maps Algorithm), it was possible to automate the process of identifying the flow regime from the differential pressure signal. The relationship between the probability density function and the power spectral density of normalized differential pressure with the gas and liquid flow rates in air-water pipe flow was also established and a machine learning algorithm (using Independent Component Analysis and Artificial Neural Networks) was proposed for the estimation of the phase flow rates from these properties. The proposed methods were adapted for use with single and dual conductivity wire-mesh sensors in vertical upwards and downwards air--water flows. A thorough evaluation of the performance and measurement uncertainty of wire-mesh sensors in gas-liquid flows was also performed. Lastly, measurements of the flow distribution in feeder tubes supplied with air-water mixtures by a simplified header model were collected and correlated to the observed flow patterns in the header.

Presented master's thesis deals with detection of brain wakefulness from scalp EEG data with higher order statistics. Part of the thesis is a description of electroencephalography, from the method of signal generation, sensing, electroencephraphy, EEG signal artifacts, frequency bands of EEG signal to its possible processing. Furthermore, the concept of mental fatigue and the possibility of its detection in the EEG signal is described. Subsequently, the principles of higher statistical methods of PCA and ICA and the specific possibilities of decomposition of EEG signal are described using these methods, from which the method of group spatial-frequency ICA was chosen as a suitable method for selection of partial oscillatory sources in EEG signal. In the next part there is described a method of acquisition of data, a the suggestion of solution with selected method and a description of the implemented algorithm, that was applied to real 256-lead scalp EEG data captured during a block task focused on subject allertnes. The absolute and relative power of the EEG signal was decomposed. From the achieved results, we observe that the fluctuations of the spatial frequency patterns of relative power (especially for theta and alpha bands) significantly more closely correspond with the change of reaction time and the error of the subjects performing the task. These observations appear to be relatively consistent with previously published literature, and the current study shows that spatial frequency ICA is able to blindly isolate space-frequency patterns whose fluctuations are statistically significantly correlated with parameters (reaction time, error rate) directly flowing from the given task.

This thesis deals with multi-channel methods of speech enhancement. Multichannel methods of speech enhancement use a few microphones for recording signals. From mixtures of signals, for example, individual speakers can be separated, noise should be reduced etc. with using neural networks. The task of separating speakers is known as a cocktail-party effect. The main method of solving this problem is called independent component analysis. At first there are described its theoretical foundation and presented conditions and requirements for its application. Methods of ICA try to separate the mixtures with help of searching the minimal gaussian properties of signals. For the analysis of independent components are used different mathematical properties of signals such as kurtosis and entropy. Signals, which were mixed artificially on a computer, can be relatively well separated using, for example, FastICA algorithm or ICA gradient ascent. However, difficult is situation, if we want to separate the signals created in the real recording enviroment, because the separation of speech people speaking at the same time in the real environment affects other various factors such as acoustic properties of the room, noise, delays, reflections from the walls, the position or the type of microphones, etc. Work presents aproach of independent component analysis in the frequency domain, which can successfully separate also recordings made in the real environment.

No description available.