O movimento dos lábios é um recurso visual relevante para a detecção da atividade de voz do locutor e para o reconhecimento da fala. Quando os lábios estão se movendo eles transmitem a idéia de ocorrências de diálogos (conversas ou períodos de fala) para o observador, enquanto que os períodos de silêncio podem ser representados pela ausência de movimentações dos lábios (boca fechada). Baseado nesta idéia, este trabalho foca esforços para detectar a movimentação de lábios e usá-la para realizar a detecção de atividade de voz. Primeiramente, é realizada a detecção de pele e a detecção de face para reduzir a área de extração dos lábios, sendo que as regiões mais prováveis de serem lábios são computadas usando a abordagem Bayesiana dentro da área delimitada. Então, a pré-segmentação dos lábios é obtida pela limiarização da região das probabilidades calculadas. A seguir, é localizada a região da boca pelo resultado obtido na pré-segmentação dos lábios, ou seja, alguns pixels que não são de lábios e foram detectados são eliminados, e em seguida são aplicados algumas operações morfológicas para incluir alguns pixels labiais e não labiais em torno da boca. Então, uma nova segmentação de lábios é realizada sobre a região da boca depois de aplicada uma transformação de cores para realçar a região a ser segmentada. Após a segmentação, é aplicado o fechamento das lacunas internas dos lábios segmentados. Finalmente, o movimento temporal dos lábios é explorado usando o modelo das cadeias ocultas de Markov (HMMs) para detectar as prováveis ocorrências de atividades de fala dentro de uma janela temporal. / Lips motion are relevant visual feature for detecting the voice active of speaker and speech recognition. When the lips are moving, they carries an idea of occurrence of dialogues (talk) or periods of speeches to the watcher, whereas the periods of silences may be represented by the absence of lips motion (mouth closed). Based on this idea, this work focus efforts to obtain the lips motion as features and to perform visual voice activity detection. First, the algorithm performs skin segmentation and face detection to reduce the search area for lip extraction, and the most likely lip regions are computed using a Bayesian approach within the delimited area. Then, the pre-segmentation of the lips is obtained by thresholding the calculated probability region. After, it is localized the mouth region by resulted obtained in pre-segmentation of the lips, i.e., some nonlips pixels detected are eliminated, and it are applied a simple morphological operators to include some lips pixels and non-lips around the mouth. Thus, a new segmentation of lips is performed over mouth region after transformation of color to enhance the region to be segmented. And, is applied the closing of gaps internal of lips segmented. Finally, the temporal motion of the lips is explored using Hidden Markov Models (HMMs) to detect the likely occurrence of active speech within a temporal window.

O movimento dos lábios é um recurso visual relevante para a detecção da atividade de voz do locutor e para o reconhecimento da fala. Quando os lábios estão se movendo eles transmitem a idéia de ocorrências de diálogos (conversas ou períodos de fala) para o observador, enquanto que os períodos de silêncio podem ser representados pela ausência de movimentações dos lábios (boca fechada). Baseado nesta idéia, este trabalho foca esforços para detectar a movimentação de lábios e usá-la para realizar a detecção de atividade de voz. Primeiramente, é realizada a detecção de pele e a detecção de face para reduzir a área de extração dos lábios, sendo que as regiões mais prováveis de serem lábios são computadas usando a abordagem Bayesiana dentro da área delimitada. Então, a pré-segmentação dos lábios é obtida pela limiarização da região das probabilidades calculadas. A seguir, é localizada a região da boca pelo resultado obtido na pré-segmentação dos lábios, ou seja, alguns pixels que não são de lábios e foram detectados são eliminados, e em seguida são aplicados algumas operações morfológicas para incluir alguns pixels labiais e não labiais em torno da boca. Então, uma nova segmentação de lábios é realizada sobre a região da boca depois de aplicada uma transformação de cores para realçar a região a ser segmentada. Após a segmentação, é aplicado o fechamento das lacunas internas dos lábios segmentados. Finalmente, o movimento temporal dos lábios é explorado usando o modelo das cadeias ocultas de Markov (HMMs) para detectar as prováveis ocorrências de atividades de fala dentro de uma janela temporal. / Lips motion are relevant visual feature for detecting the voice active of speaker and speech recognition. When the lips are moving, they carries an idea of occurrence of dialogues (talk) or periods of speeches to the watcher, whereas the periods of silences may be represented by the absence of lips motion (mouth closed). Based on this idea, this work focus efforts to obtain the lips motion as features and to perform visual voice activity detection. First, the algorithm performs skin segmentation and face detection to reduce the search area for lip extraction, and the most likely lip regions are computed using a Bayesian approach within the delimited area. Then, the pre-segmentation of the lips is obtained by thresholding the calculated probability region. After, it is localized the mouth region by resulted obtained in pre-segmentation of the lips, i.e., some nonlips pixels detected are eliminated, and it are applied a simple morphological operators to include some lips pixels and non-lips around the mouth. Thus, a new segmentation of lips is performed over mouth region after transformation of color to enhance the region to be segmented. And, is applied the closing of gaps internal of lips segmented. Finally, the temporal motion of the lips is explored using Hidden Markov Models (HMMs) to detect the likely occurrence of active speech within a temporal window.

La tomographie en 2D et 3D sont largement utilisée dans l’imagerie médicale ainsi que dans le Contrôle Non Destructif (CND) pour l’industrie. Dans toutes les deux applications, il est nécessaire de réduire le nombre de projections. Dans certains cas, la reconstruction doit être faite avec un nombre d’angle de projections limité. Les données mesurées sont toujours avec des erreurs (erreurs de mesure et de modélisation). Nous sommes donc presque toujours dans la situation de problèmes inversés mal posés. Le rôle des méthodes probabilistes et de la modélisation a priori devient crucial. Pour la modélisation a priori, en particulier dans les applications NDT, l’objet à l’examen est composé de plusieurs matériaux homogènes, avec plusieurs blocs continus séparés par des discontinuités et des contours. Ce type d’objet est dit continu par morceaux. L’objet de cette thèse est sur la reconstruction des objets continu ou constante par morceaux, ou plus généralement homogène par morceaux. En résumé, deux méthodes principales sont proposées dans le contexte de l’inférence bayésienne. La première méthode consiste à reconstruire l’objet en imposant que sa transformée de Haar soit parcimonieuse. Un modèle bayésien hiérarchique est proposé. Dans cette méthode, les variables et les paramètres sont estimés et les hyper-paramètres sont initialisés selon la définition des modèles antérieurs. La deuxième méthode reconstruit les objets en estimant simultanément les contours. L’objet continu par morceaux est modélisé par un modèle markovien non-homogène, qui dépend du gradient de l’objet, et le gradient dépend aussi de l’estimation de l’objet. Cette méthode est également semi-supervisé, avec les paramètres estimés automatiquement. Ces méthodes sont adaptées aux reconstructions de grande taille de données 3D, dans lesquelles le processeur GPU est utilisé pour accélérer les calculs. Les méthodes sont validées avec des données simulées et des données réelles, et sont comparées avec plusieurs méthodes classiques. / 2D and 3D X-ray Computed Tomography (CT) is widely used in medical imaging as well as in Non Destructive Testing (NDT) for industrial applications. In both domains, there is a need to reduce the number of projections. In some cases we may also be limited in angles. The measured data are always with errors (measurement and modelling errors). We are consequently almost always in the situation of ill-posed inverse problems. The role of the probabilistic methods and the prior modelling become crucial. For prior modelling, in particular in NDT applications, the object under examination is composed with several homogeneous materials, with several continuous blocs separated by some discontinuities and contours. This type of object is called the piecewise-continuous object. The focus of this thesis on the reconstruction of the picewise continuous or constant, or more generally piecewise homogeneous objects. In summary two main methods are proposed in the context of the Bayesian inference. The first method consists in reconstructing the object while enforcing the sparsity of the discrete Haar transformation coefficients of the object. A hierarchical Bayesian model is proposed. In this method, the unknown variables and parameters are estimated and the hyper-parameters are initialized according to the definition of prior models. The second method reconstruct objects while the contours are estimated simultaneously. The piecewise continuous object is modeled by a non-homogeneous Markovian model, which depends on the gradient of the object, while the gradient also depends on the estimation of the object. In this methods, the semi-supervised system model is also achieved, with the parameters estimated automatically. Both methods are adapted to the 3D big data size reconstructions, in which the GPU processor is used to accelerate the computation. The methods are validated with both simulated and real data, and are compared with several conventional state-of-the-art methods.

Este texto trata da estimação de espectro direcional de ondas a partir de movimentos de primeira ordem de sistemas oceânicos estacionários utilizando um método de inferência bayesiano. Após descrição do método e discutidas suas principais fontes de incertezas, este foi validado experimentalmente através de ensaios com dois tipos de embarcações para ondas irregulares e mares bimodais, e verificado com dados de monitoração em escala real. Uma comparação preliminar com dados provenientes de um radar próximo ao sistema monitorado também foi feita. Com base nestes resultados, discute-se a eficiência do método e suas limitações. / This text is about estimating directional wave spectra based on first order motions of stationary systems by means of a bayesian inference method. After being described the method and its main sources of uncertainties discussed, an experimental validation was carried out with two different types of hulls for irregular waves and bimodal seas, with a posterior verification using data from a real scale monitoring campaign. Also, a preliminary comparison with data from a radar in the vicinities of the monitored system was made. Based on those results, the methods efficiency is discussed, along with its limitations.

Made available in DSpace on 2016-04-27T16:57:27Z (GMT). No. of bitstreams: 1 Pericles Cesar de Araujo.pdf: 2815289 bytes, checksum: 8b5efd754b4e27e42558c8f0ebe0c75e (MD5) Previous issue date: 2013-08-18 / This thesis aims at dealing with issues concerning the scientification of Mathematics Education and, consequently, the methods used in researches in that area. It means dealing with methodological issues, under qualitative, quantitative and mixed perspectives with the objective of discussing the increase, always desired, of the reliability of data analysis. It is a theoretical research, with methodological procedures adequate to this kind of study, that is, bibliographical review, analysis of thesis, dissertations, articles and books. Among the theoretical background, we mention Didactic Engineering, Conceptual Metaphor, Fuzzy Logics, Bayesian Statistics and elements from Popper, Kuhn and Lakatos theories. The Didactic Engineering was considered the common thread of reflections; the Fuzzy Logics, inserted for the research universe in Mathematics Education, is characterized by an important heterogeneity of the phenomena. The Bayesian Statistics is a referential to deal with the qualitative aspect. As results, we propose in this thesis, mixed methods of research in Mathematics Education. The first is the one that considers the combination of two methods: the Bayesian method and the Didactic Engineering; the second is the Fuzzy and Bayesian. This way, we understand that with this thesis, we contribute to the discussion on the use of mixed methodology in researches in Mathematics Education, as well as present a method with potential to develop the research in Mathematics Education with scientific characteristics that meet the demands of Popper, Kuhn and Lakatos / Esta tese tem por objetivo tratar das questões relativas à cientificidade da Educação Matemática, e em consequência, dos métodos utilizados na pesquisa dessa área. Ou seja, tratar das questões metodológicas, nas perspectivas qualitativas, quantitativas ou mistas com vistas a discutir a ampliação, sempre desejável, da confiabilidade da análise de dados. Trata-se de uma pesquisa de caráter teórico, com procedimentos metodológicos adequados a esse tipo de estudo, ou seja, realização de levantamento bibliográfico, análise de dissertações, teses, artigos, e livros. Entre as bases teóricas, estão a Engenharia Didática Clássica, a Metáfora Conceitual, a Lógica Difusa, a Estatística Bayesiana, e elementos das teorias de Popper, Kuhn e Lakatos. A Engenharia Didática Clássica foi considerada um fio condutor das reflexões; a Lógica Difusa é inserida, pois o universo da pesquisa em Educação Matemática é caracterizado por uma acentuada heterogeneidade de fenômenos. A Estatística Bayesiana é um referencial para tratar do aspecto quantitativo. Como resultado são propostos nesta tese métodos mistos de pesquisas em Educação Matemática. O primeiro deles é aquele que considera a agregação de dois métodos: o Método Bayesiano e a Engenharia Didática Clássica; o segundo é o Método Estatístico Bayesiano de Dados Difusos. Assim, avaliamos que trazemos, com esta tese, uma contribuição para a discussão sobre o uso de metodologias mistas na pesquisa em Educação Matemática, bem como que apresentamos um método com potencialidades de munir a pesquisa em Educação Matemática de características científicas que atendam às exigências de Popper, Kuhn e Lakatos

Πέραν από τις κλασσικές μεθόδους ηλεκτρονικής μάθησης που εφαρμόζονται σήμερα, προτείνεται ένας συνδυασμός εξελικτικών αλγορίθμων και τεχνητής νοημοσύνης για την δημιουργία έξυπνων προσαρμοστικών συστημάτων ηλεκτρονικής μάθησης. Σε αυτή τη διπλωματική εργασία περιγράφονται και παρουσιάζονται οι προτεινόμενοι αλγόριθμοι και ταυτόχρονα η προτεινόμενη πιλοτική εφαρμογή. Το προτεινόμενο σύστημα μπορεί να προβλέψει τις μαθησιακές ικανότητες του μαθητή, μέσα από εξεταστικές διαδικασίες οι οποίες προσφέρονται από το σύστημα, με αποτέλεσμα, το σύστημα να είναι σε θέση να προβλέψει τις επόμενες κινήσεις του μαθητή. Μέσα από την προτεινόμενη εφαρμογή αναπτύχθηκαν μηχανισμοί οι οποίοι συλλέγουν πληροφορίες για τον κάθε χρήστη ξεχωριστά και δημιουργούν ανεξάρτητα προφίλ χρήστη για τον κάθε ένα. Με την χρήση συνδυασμού εξελικτικών αλγορίθμων και αλγορίθμων μάθησης το σύστημα εκπαιδεύεται ώστε να μπορεί να προβλέπει τις μελλοντικές κινήσεις του χρήστη. Η εφαρμογή που αναπτύχτηκε είναι βασισμένη σε τεχνολογίες διαδικτύου, βάσεις δεδομένων και τεχνολογίες έξυπνων πρακτόρων. / As a step beyond the classic e-learning methods that are applied today, the combination of evolutionary programming with artificial intelligence has incorporated in order to create an intelligent adaptive e-learning system. In this thesis the theory of the proposed algorithms are presented and the proposed pilot application too. The proposed system can predict the learning possibilities of a student, concerning the knowledge that is provided to him by the system, thus providing the ability to the machine to predict and anticipate his reactions. We have developed applications that can collect information for the student’s history, thus creating concrete individual profiles. Then, using evolutionary programming techniques combined with machine learning algorithms the system is trained in order to can henceforth calculate and anticipate the student’s knowledge. The applications that have been developed are based on internet technologies, data bases and intelligent agents’ technology.

Tandem mass spectrometry (MS/MS) is the dominant approach for large-scale peptide sequencing in high-throughput proteomic profiling studies. The computational analysis of MS/MS spectra involves the identification of peptides from experimental spectra, especially those with post-translational modifications (PTMs), as well as the inference of protein composition based on the putative identified peptides. In this thesis, we tackled two major challenges associated with an MS/MS analysis: 1) the refinement of PTM predictions from MS/MS spectra and 2) the inference of protein composition based on peptide predictions. We proposed two PTM prediction refinement algorithms, PTMClust and its Bayesian nonparametric extension \emph{i}PTMClust, and a protein identification algorithm, pro-HAP, that is based on a novel two-layer hierarchical clustering approach that leverages prior knowledge about protein function. Individually, we show that our two PTM refinement algorithms outperform the state-of-the-art algorithms and our protein identification algorithm performs at par with the state of the art. Collectively, as a demonstration of our end-to-end MS/MS computational analysis of a human chromatin protein complex study, we show that our analysis pipeline can find high confidence putative novel protein complex members. Moreover, it can provide valuable insights into the formation and regulation of protein complexes by detailing the specificity of different PTMs for the members in each complex.

Tandem mass spectrometry (MS/MS) is the dominant approach for large-scale peptide sequencing in high-throughput proteomic profiling studies. The computational analysis of MS/MS spectra involves the identification of peptides from experimental spectra, especially those with post-translational modifications (PTMs), as well as the inference of protein composition based on the putative identified peptides. In this thesis, we tackled two major challenges associated with an MS/MS analysis: 1) the refinement of PTM predictions from MS/MS spectra and 2) the inference of protein composition based on peptide predictions. We proposed two PTM prediction refinement algorithms, PTMClust and its Bayesian nonparametric extension \emph{i}PTMClust, and a protein identification algorithm, pro-HAP, that is based on a novel two-layer hierarchical clustering approach that leverages prior knowledge about protein function. Individually, we show that our two PTM refinement algorithms outperform the state-of-the-art algorithms and our protein identification algorithm performs at par with the state of the art. Collectively, as a demonstration of our end-to-end MS/MS computational analysis of a human chromatin protein complex study, we show that our analysis pipeline can find high confidence putative novel protein complex members. Moreover, it can provide valuable insights into the formation and regulation of protein complexes by detailing the specificity of different PTMs for the members in each complex.

Este texto trata da estimação de espectro direcional de ondas a partir de movimentos de primeira ordem de sistemas oceânicos estacionários utilizando um método de inferência bayesiano. Após descrição do método e discutidas suas principais fontes de incertezas, este foi validado experimentalmente através de ensaios com dois tipos de embarcações para ondas irregulares e mares bimodais, e verificado com dados de monitoração em escala real. Uma comparação preliminar com dados provenientes de um radar próximo ao sistema monitorado também foi feita. Com base nestes resultados, discute-se a eficiência do método e suas limitações. / This text is about estimating directional wave spectra based on first order motions of stationary systems by means of a bayesian inference method. After being described the method and its main sources of uncertainties discussed, an experimental validation was carried out with two different types of hulls for irregular waves and bimodal seas, with a posterior verification using data from a real scale monitoring campaign. Also, a preliminary comparison with data from a radar in the vicinities of the monitored system was made. Based on those results, the methods efficiency is discussed, along with its limitations.

Submitted by (ana.araujo@ufrpe.br) on 2016-08-09T12:57:36Z No. of bitstreams: 1 Niedja Maristone Oliveira Barreto Queiroz.pdf: 2909360 bytes, checksum: 109caf21db04442310458a38ed638100 (MD5) / Made available in DSpace on 2016-08-09T12:57:36Z (GMT). No. of bitstreams: 1 Niedja Maristone Oliveira Barreto Queiroz.pdf: 2909360 bytes, checksum: 109caf21db04442310458a38ed638100 (MD5) Previous issue date: 2004-03-26 / Logistic regression has application in several fields as epidemiology, medical research, banks, market research and social research. One of its advantages is that the interpretation of the measure is possible through the " Odds Ratios” (OR), that are functions of the parameters of the model. In this study the binary regression model was used, with the objective of estimating the relationship between two variables, taking into account the presence of other factors. For his purpose a Bayesian approach was used to estimate those risk measures, and these results were compared with the corresponding classical results obtained by application of a stepwise backward process, using the maximum likelihood as criterion for exclusion of the variable of the model, and the Wald test as analysis of each parameter of the final model, both at the level of significance of 0,05. An application was performed using real data from a transverse study of 106.203 blood donor candidates, found apt by the clinical screening process performed at the blood bank Recife of the HEMOPE foundation. Measures of HIV infection association “OR” were estimated in relation with certain socio-demographic conditions, sorological markers for other Sexually Transmissible Diseases as well as the donation type. For the classical analysis thestatistical package SPSS version 10 was used, and for the bayesian analysis the Winbugs 14. The results indicated that OR obtained using the two methods are rather similar, in spite of the fact that the classical approach used Maximum likelihood and the bayesian approach used the Markov Chain Monte Carlo(MCMC), which are quite different methods. It was concluded, that the factors independently associated to the HIV infection risk among donors of blood in the observed period, for the bayesian estimate, were: age 18 to 28 years (2,45) and 29 to 39 years (2,79); illiteracy (8,17), primary school (3,31) and secundary school (3,29); positive Anti-Hbc (1,95), positive syphilis (3,14), residence in the Metropolitan Area of Recife (2,41) and type of voluntary donation (11,94). / Regressão logística tem aplicação em diversos campos como epidemiologia, pesquisa médica, bancos, pesquisa de mercado e pesquisa social. Uma de suas vantagens é que a interpretação da medida é possível através das “Odds Ratios” (OR), que são funções dos parâmetros do modelo. Neste estudo foi usado o modelo de regressão binária, com o objetivo de estimar a relação entre duas variáveis tendo em conta a presença de outros fatores. Utilizou-se para isso uma abordagem bayesiana para estimar essas medidas de risco, fazendo uma comparação com os resultados da abordagem clássica proveniente de um processo stepwise backward, utilizando o critério da razão de verossimilhança como exclusão da variável do modelo e o teste de Wald como análise de cada parâmetro do modelo final, ambos no nível de significância de 0,05. Realizou-se uma aplicação com dados reais proveniente de um estudo transversal de 106.203 doadores de sangue de 1ª doação aptos na triagem clínica no Hemocentro Recife da Fundação HEMOPE. Estimou-se medidas de associação “OR”, da infecção por HIV, com relação a algumas condições sócio-demográficas, marcadores sorológicos para outras Doenças Sexualmente Transmissíveis (DST) e tipo de doação. Para as análises no método clássico foi utilizado o pacote estatístico SPSS versão 10 e no método bayesiano o Winbugs 14. Os resultados indicaram que as OR estimadas, utilizando os dois métodos, foram bastante próximas, apesar do clássico utilizar o método de estimação por Máxima Verossimilhança, e o bayesiano utilizar os métodos de Monte Carlo Cadeia de Markov (MCMC), que são métodos diferentes. Concluiu-se, que os fatores independentemente associados ao risco de infecção por HIV entre doadores de sangue no período foram, pela estimativa bayesiana: idade 18 a 28 anos (2,45) e 29 a 39 anos (2,79); escolaridade: analfabeto (8,17), ensino fundamental (3,31) e médio (3,29); Anti-Hbc positivo (1,95); sífilis positivo (3,14); residir na Região Metropolitana do Recife (RMR) (2,41) e tipo de doação voluntária (11,94).