This dissertation is concerned with the problem of describing the singing voice within the audio signal of a song. This work is motivated by the fact that the lead vocal is the element that attracts the attention of most listeners. For this reason it is common for music listeners to organize and browse music collections using information related to the singing voice such as the singer name. Our research concentrates on the three major problems of music information retrieval: the localization of the source to be described (i.e. the recognition of the elements corresponding to the singing voice in the signal of a mixture of instruments), the search of pertinent features to describe the singing voice, and finally the development of pattern recognition methods based on these features to identify the singer. For this purpose we propose a set of novel features computed on the temporal variations of the fundamental frequency of the sung melody. These features, which aim to describe the vibrato and the portamento, are obtained with the aid of a dedicated model. In practice, these features are computed on the time-varying frequency of partials obtained using the sinusoidal model. In the first experiment we show that partials corresponding to the singing voice can be accurately differentiated from the partials produced by other instruments using decisions based on the parameters of the vibrato and the portamento. Once the partials emitted by the singer are identified, the segments of the song containing singing can be directly localized. To improve the recognition of the partials emitted by the singer we propose to group partials that are related harmonically. Partials are clustered according to their degree of similarity. This similarity is computed using a set of CASA cues including their temporal frequency variations (i.e. the vibrato and the portamento). The clusters of harmonically related partials corresponding to the singing voice are identified using the vocal vibrato and the portamento parameters. Groups of vocal partials can then be re-synthesized to isolate the voice. The result of the partial grouping can also be used to transcribe the sung melody. We then propose to go further with these features and study if the vibrato and portamento characteristics can be considered as a part of the singers' signature. Previous works on singer identification describe audio signals using features extracted on the short-term amplitude spectrum. The latter features aim to characterize the timbre of the sound, which, in the case of singing, is related to the vocal tract of the singer. The features we develop in this document capture long-term information related to the intonation of the singer, which is relevant to the style and the technique of the singer. We propose a method to combine these two complementary descriptions of the singing voice to increase the recognition rate of singer identification. In addition we evaluate the robustness of each type of feature against a set of variations. We show the singing voice is a highly variable instrument. To obtain a representative model of a singer's voice it is thus necessary to build models using a large set of examples covering the full tessitura of a singer. In addition, we show that features extracted directly from the partials are more robust to the presence of an instrumental accompaniment than features derived from the amplitude spectrum.

Made available in DSpace on 2014-12-17T14:08:38Z (GMT). No. of bitstreams: 1 AuciomarCTC_DISSERT.pdf: 1704122 bytes, checksum: fe303bbcf110f6668682831da2f3480d (MD5) Previous issue date: 2009-12-16 / The oscillations presents in control loops can cause damages in petrochemical industry. Canceling, or even preventing such oscillations, would save up to large amount of dollars. Studies have identified that one of the causes of these oscillations are the nonlinearities present on industrial process actuators. This study has the objective to develop a methodology for removal of the harmful effects of nonlinearities. Will be proposed an parameter estimation method to Hammerstein model, whose nonlinearity is represented by dead-zone or backlash. The estimated parameters will be used to construct inverse models of compensation. A simulated level system was used as a test platform. The valve that controls inflow has a nonlinearity. Results and describing function analysis show an improvement on system response / As oscila??es presentes nas malhas de controle podem causar preju?zos ? ind?stria petroqu?mica. Anular, ou at? mesmo amenizar tais oscila??es, acarretaria em uma grande economia. Estudos identificaram que um dos causadores destas oscila??es s?o as n?o linearidades presentes nos atuadores dos processos industriais. O presente trabalho tem como objetivo desenvolver uma metodologia para cancelamento dos efeitos nocivos destas n?o linearidades. Ser? proposto um m?todo de estima??o de par?metros para o modelo de Hammerstein, cuja n?o linearidade ? representada pela zona-morta ou pela folga. Os par?metros estimados ser?o usados na constru??o de modelos inversos de compensa??o. Utilizou-se um sistema de n?vel simulado como plataforma de testes, o qual possui uma v?lvula de controle de fluxo da vaz?o de entrada com uma n?o linearidade associada. Resultados e a an?lise por fun??o descritiva mostram a melhoria do comportamento da sa?da do processo

Ce travail de thèse propose une approche générique pour l'estimation de l'état/ des paramètres et pour le placement de capteurs de systèmes hyperboliques non linéaires en dimension infinie. Le travail est donc divisé en deux parties principales : une partie consacrée à l'estimation optimale et une partie dédiée au placement optimal de capteurs. La méthode d'estimation optimale utilise une approche par calcul des variations et utilise la méthode des multiplicateurs de Lagrange. Ces multiplicateurs jouent un rôle important en donnant accès aux sensibilités des mesures par rapport aux variables qui doivent être estimées. Ces sensibilités, décrites par les équations adjointes, sont aussi à l'origine d'une nouvelle approche, dite méthode de l'adjoint, pour le placement optimal de capteurs. Divers exemples, construits sur la base de simulations mais également de données réelles et pour différents scénarios, sont aussi étudiées afin d'illustrer l'efficacité des approches développées. Ces exemples concernent les écoulements à surface libre (en hydrologie des bassins versants) et le trafic routier représentés par des équations aux dérivées partielles hyperboliques non linéaires. / The thesis proposes a general framework for both state/parameters estimation and sensor placement in nonlinear infinite dimensional hyperbolic systems. The work is therefore divided into two main parts: a first part devoted to the optimal estimation and a second one to optimal sensor location. The estimation method is based on the calculus of variations and the use of Lagrange multipliers. The Lagrange multipliers play an important role in giving access to the sensitivities of the measurements with respect to the variables to be estimated. These sensitivities, described by the adjoint equations, are also the key idea of a new approach, so-called the adjoint-based approach, for the optimal sensor placement. Various examples, either based on some simulations with synthetic measurements or real data sets and for different scenarios, are also studied to illustrate the effectiveness of the developed approaches. Theses examples concern the overland flow systems and the traffic flow, which are both governed by nonlinear hyperbolic partial differential equations.

Le travail présenté dans cette thèse a pour objectif d'apporter une contribution aux méthodes de commande et d'observation des machines asynchrones destinées à la traction électrique. Dans ce contexte, plusieurs algorithmes ont été développés et implémentés. Après une présentation rapide de la commande vectorielle classique, de nouvelles approches de commande non linéaire sont proposées : il s'agit plus précisément de la commande backstepping classique et sa variante avec action intégrale. Une deuxième partie est consacrée à l'observation et à l'estimation des paramètres et des états de la machine, basée sur des structures MRAS-modes glissants d'une part et sur des structures de filtrage synchrone d'autre part. Une analyse détaillée du problème de fonctionnement à basse vitesse nous a conduit à proposer une solution originale dans le cadre d'une commande sans capteur mécanique. Le problème de la dégradation du couple en survitesse a été traité par un algorithme de défluxage basé sur la conception d'un contrôleur de tension. Enfin, nous avons proposé un algorithme d'optimisation afin de minimiser les pertes dans l'ensemble Onduleur-Machine / The work presented in this thesis aims to contribute to the control and observation of the induction machines for electric traction. Several algorithms have been developed and implemented. After a fast presentation of the classical vector control, new approaches of non-linear control are proposed : the classical backstepping and integral backstepping. A second part deals with the observation and the estimation of parameters and states of the machine, based on MRAS-Sliding Mode structures on one hand and on synchronous filtering structures on the other hand. A detailed analysis of the operation at low speed led us to propose an original solution for a Sensorless control. The torque degradation in field weakening zone was treated by a voltage regulation controller. Finally, we proposed losses minimization algorithm for the Inverter-Machine set

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Nesta tese é realizada a modelagem do comportamento hidráulico dos principais rios que compõem a bacia hidrográfica do Rio Bengalas, localizada no município de Nova Friburgo-RJ, a qual abrange a área mais urbanizada da referida cidade. Para a realização das simulações foi utilizado o Sistema de Modelagem de Águas MOHID, ferramenta MOHID Land. Já para a calibração do modelo foram adotados alguns métodos de otimização, mais precisamente, os algoritmos de Luus- Jaakola (LJ) e Colisão de Partículas (PCA), acoplados ao referido sistema, com o intuito de determinar os principais parâmetros necessários à modelagem de corpos hídricos, bem como suas bacias hidrográficas. Foram utilizados dados topográficos do IBGE disponibilizados pela prefeitura após a elaboração do Plano de Águas Pluviais da região de interesse. Com o modelo devidamente calibrado por meio de dados experimentais, foi realizada a validação do mesmo através da simulação de inundações nesta região. Apesar de técnicas de otimização acopladas à plataforma MOHID terem sido utilizadas pela primeira vez em um rio de montanha, os resultados apresentaram-se importantes e qualitativamente satisfatórios do ponto de vista de auxílio à tomada de decisões, tendo como base a prevenção de danos causados pelas elevações da lâmina dágua que ocorrem frequentemente em Nova Friburgo, como por exemplo, a recente tragédia de janeiro de 2011 ocorrida na Região Serrana do Estado do Rio de Janeiro. / In this thesis is held the modeling of the hydraulic behavior of the main rivers that form the Bengalas River watershed, located in the municipality of Nova Friburgo, Rio de Janeiro, which covers the most urbanized area of that city. To perform the simulations was used MOHID Waters Modeling System, Land MOHID tool. As for the calibration of the model were adopted some optimization methods, more precisely, the Luus-Jaakola (LJ) and Particle Collision (PCA) algorithms coupled to said system, in order to determine the main parameters needed for modeling bodies water and their watersheds. We used topographic data from the IBGE provided by the city after the preparation of the Stormwater Plan for the region of interest. With the model calibrated through experimental data, the same was validated by simulating flooding in this region. Although optimization techniques coupled to the platform MOHID were first used in a mountain river, the results presented are important and qualitatively satisfactory from the point of view of aid to decision making, based on the prevention of damage caused by increases in water level that often occur in Nova Friburgo, such as the recent tragedy that occurred in January 2011 in the mountainous region of the Rio de Janeiro State.

Made available in DSpace on 2016-08-17T18:39:30Z (GMT). No. of bitstreams: 1 2145.pdf: 2206472 bytes, checksum: 5295597725f34bdf5560d6cda8af7446 (MD5) Previous issue date: 2008-03-27 / Financiadora de Estudos e Projetos / Streptococcus pneumoniae (pneumococo) is a pathogenic bacterium that causes several infections which are aggravated by the increase of serotypes with antibiotics resistance. The development of an effective vaccine against this pathogen is crucial for the prevention of the neumococcal illnesses. Conjugated vaccines, consisting of the capsular polysaccharide joined to a carrier protein, are more efficient in the stimulation of the immunologic memory. The capsular polysaccharide (PS) is present in the capsule that involves the cell. Thus, the conjugated vaccine elaboration involves bacterial cells cultivation for its production. As the organism is cultivated in the oxygen absence, the lactate production is inevitably high, leading to growth inhibition due to lactate accumulation in the medium. To minimize the inhibitory effects of the lactate accumulation and to increase the PS production it is necessary to monitor the process and adequately control the addition of supplementary medium along with the withdrawal of saturated medium. This kind of operation can be performed by carrying out a fed-bath cultivation in a bioreactor connected to a perfusion system. The success on the monitoring, control and optimization of this bioprocess depends on the efficiency of the modeling and simulation resources employed. This research work considers the uses intelligent computational techniques, specifically the technique of heuristical search called simulated annealing (SA) combined with neural networks for the state inference and the optimization of S. pneumoniae cultivations. The proposal was implemented as a computational system that: a) uses the SA for the identification of the values for a set of parameters associated to unstructured models and; b) uses neural networks (individually and grouped as a committee) for the state inference of a culture. The work presents and discusses the results of the system for data sets experimentally obtained and highlights the importance of the proposal for achieving a higher efficiency in the culture control processes. / Streptococcus pneumoniae (pneumococo) é uma bactéria patogênica causadora de várias infecções que são agravadas pelo aumento de cepas com resistência aos antibióticos. O desenvolvimento de uma vacina efetiva contra este patógeno é crucial para a prevenção das doenças pneumocócicas. Vacinas conjugadas, constituídas pelo polissacarídeo capsular ligado a uma proteína carregadora, são mais eficientes no estímulo da memória imunológica. O polissacarídeo capsular (PS) está presente na cápsula que envolve a célula e, desta forma, a elaboração de vacinas conjugadas envolve o cultivo da bactéria para a produção do mesmo. Como o microrganismo é cultivado na ausência de oxigênio, a produção de lactato é inevitavelmente elevada e o seu acúmulo no meio provoca a inibição do crescimento. Para minimizar os efeitos inibitórios da acumulação de lactato e aumentar a produção de PS é necessário monitorar o processo e controlar adequadamente a adição de meio suplementar e a retirada de meio saturado em cultivos operados em batelada alimentada, utilizando biorreatores acoplados a sistema de perfusão. O sucesso no monitoramento, no controle e na otimização deste bioprocesso depende da utilização de recursos de modelagem e de simulação que sejam eficientes. Este trabalho de pesquisa propõe o uso de técnicas computacionais inteligentes, especificamente a técnica de busca heurística chamada de simulated annealing (SA) aliada a redes neurais, para a inferência de estado e a otimização de cultivos de S. pneumoniae. A proposta foi concretizada via desenvolvimento de um sistema computacional que: a) faz uso do SA para a identificação do conjunto de valores de parâmetros associados a modelos não estruturados e; b) usa redes neurais (individualmente e em regime de comitê) para a inferência de estado de um cultivo. O trabalho apresenta e discute os resultados do sistema em conjuntos de dados obtidos experimentalmente e evidencia a importância da proposta para uma maior eficiência no controle de processos de cultivo.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Nesta tese é realizada a modelagem do comportamento hidráulico dos principais rios que compõem a bacia hidrográfica do Rio Bengalas, localizada no município de Nova Friburgo-RJ, a qual abrange a área mais urbanizada da referida cidade. Para a realização das simulações foi utilizado o Sistema de Modelagem de Águas MOHID, ferramenta MOHID Land. Já para a calibração do modelo foram adotados alguns métodos de otimização, mais precisamente, os algoritmos de Luus- Jaakola (LJ) e Colisão de Partículas (PCA), acoplados ao referido sistema, com o intuito de determinar os principais parâmetros necessários à modelagem de corpos hídricos, bem como suas bacias hidrográficas. Foram utilizados dados topográficos do IBGE disponibilizados pela prefeitura após a elaboração do Plano de Águas Pluviais da região de interesse. Com o modelo devidamente calibrado por meio de dados experimentais, foi realizada a validação do mesmo através da simulação de inundações nesta região. Apesar de técnicas de otimização acopladas à plataforma MOHID terem sido utilizadas pela primeira vez em um rio de montanha, os resultados apresentaram-se importantes e qualitativamente satisfatórios do ponto de vista de auxílio à tomada de decisões, tendo como base a prevenção de danos causados pelas elevações da lâmina dágua que ocorrem frequentemente em Nova Friburgo, como por exemplo, a recente tragédia de janeiro de 2011 ocorrida na Região Serrana do Estado do Rio de Janeiro. / In this thesis is held the modeling of the hydraulic behavior of the main rivers that form the Bengalas River watershed, located in the municipality of Nova Friburgo, Rio de Janeiro, which covers the most urbanized area of that city. To perform the simulations was used MOHID Waters Modeling System, Land MOHID tool. As for the calibration of the model were adopted some optimization methods, more precisely, the Luus-Jaakola (LJ) and Particle Collision (PCA) algorithms coupled to said system, in order to determine the main parameters needed for modeling bodies water and their watersheds. We used topographic data from the IBGE provided by the city after the preparation of the Stormwater Plan for the region of interest. With the model calibrated through experimental data, the same was validated by simulating flooding in this region. Although optimization techniques coupled to the platform MOHID were first used in a mountain river, the results presented are important and qualitatively satisfactory from the point of view of aid to decision making, based on the prevention of damage caused by increases in water level that often occur in Nova Friburgo, such as the recent tragedy that occurred in January 2011 in the mountainous region of the Rio de Janeiro State.

Made available in DSpace on 2016-08-17T14:53:03Z (GMT). No. of bitstreams: 1 Jorge Farid Amate.pdf: 1076998 bytes, checksum: 7c54d6bc2833288f1945368aa6976936 (MD5) Previous issue date: 2009-02-06 / Conselho Nacional de Desenvolvimento Científico e Tecnológico / In industrial processes, where parameters estimation and standard recognition are desired, digital filters technology is used to do estimation. The digital filter is responsible for prediction and filtering process. Then, the filter behavior can be analyzed based on performance, gains and others variables linked to the specified model. However, to obtain trusty variables and data to estimate the process in question, a model that represents well the physics plant is needed. To do this, are applied techniques based on Systems Identification where we obtain the ARX, ARMAX, Output-Error and Box-Jenkins models of the electrolytic pot. Results, validation and their analysis, applied in standard recognition, using different structures are presented. / Em processos industriais, onde deseja-se a estimação de parâmetros e reconhecimento de padrões, utiliza-se da tecnologia de filtros digitais para tal fim. O filtro digital é responsável pelo processo de predição e filtragem. Assim, pode-se fazer uma análise do comportamento do filtro baseada no desempenho, ganhos e outras variáveis ligadas ao modelo especificado. Porém, para obtenção de variáveis e dados confiáveis para estimar-se o processo em questão, necessita-se de um modelo que represente bem a planta física. Para isto, são aplicadas técnicas baseadas em Identificação de Sistemas, onde são obtidos os modelos ARX, ARMAX, Output-Error e Box-Jenkins da cuba eletrolítica. São apresentados os resultados, validações dos modelos e análise dos mesmos, aplicados ao reconhecimento de padrões, utilizando-se diferentes estruturas.

[pt] Exames de ultrassom são uma ferramenta popular de aquisição de imagens na medicina atual por ser um procedimento não-invasivo, seguro e barato. Entretanto, inerente a qualquer exame de ultrassom encontra-se o ruído speckle, responsável pela degradação da imagem e dificultando tanto sua interpretação por parte de médicos e pacientes, quanto prejudicando a acurácia de métodos computacionais de pós processamento, como classificação, reconstrução, caracterização de tecidos e segmentação, entre outros. Portanto, métodos de remoção ou suavização deste ruído que preservem as principais características do conteúdo observado se fazem fundamentais para um avanço nestes processos. Definido como um ruído multiplicativo, que segue estatísticas não-Gaussianas e como fortemente correlacionado, sua solução ainda hoje é tema de debates e estudos. Neste trabalho apresentaremos diversos métodos de filtragem 2D que se propõem a reduzir ou solucionar o ruído speckle bem como métodos qualitativos para avaliar seus desempenhos e técnicas para escolher os melhores parâmetros de cada filtro a fim de eleger quais métodos melhor solucionam este ruído. / [en] Ultrasound exams are a popular tool for image acquisition in day-to-day medicine, since it is a noninvasive, safe and cheap procedure. However, speckle noise is intrinsic to any ultrasound exam, and it is responsible for image quality degradation and for hindering its interpretation by doctors and patients alike, while also impairing the accuracy of post processing computational methods, such as classification, reconstruction, tissue characterization and segmentation, among others. Hence, smoothing or denoising methods that preserves the observed content core attributes are essential for those processes. Defined as a multiplicative noise, following non-Gaussian statistics and as strongly correlated, its solution today is still a matter of debates and research. In this work, several 2D filters that aim to smooth or remove speckle noise along with qualitative methods to evaluate their performances and means of choosing their best parameters are presented.

Les travaux présentés dans le cadre de cette thèse traitent de la modélisation mathématique de la croissance tumorale. La première partie de cette thèse traite de l’estimation des paramètres. Plus précisément, il s’agit de déterminer la vascularisation d’une tumeur à partir de sa dynamique. Pour cela, nous générons à partir d’un modèle d’équations aux dérivées partielles l’évolution en temps de la densité de cellules tumorales. Ensuite, nous résolvons des problèmes inverses afin de retrouver la densité de vascularisation correspondante. Nous montrons que la vascularisation estimée permet de prédire efficacement la croissance future de la tumeur. Dans un second temps, nous introduisons une classe de modèles pour la croissance de gliomes qui sont adaptés à la fois aux gliomes de bas grades et aux glioblastomes multiformes. Afin de tenir compte des spécificités des gliomes, le modèle prend en considération le caractère infiltrant de ce type de tumeur ainsi que l’hétérogénéité, l’anisotropie et la géométrie du cerveau. Nos modèles permettent d’étudier l’efficacité des traitements anti-angiogéniques et de la comparer à celle d’un traitement qui inhiberait la capacité d’invasion de gliomes. Les modèles ont été implémentés en 2D et en 3D dans des géométries réalistes obtenues grâce à un atlas. / This thesis deals with mathematical modeling of tumor growth. Firstly, we present a parameter estimation method. More precisely, it consists in recovering the position of the tumor blood vessel, starting from imaging. The first step is to design a particular vascularization, then we compute the tumor growth with this blood-vessel network by using a model based on partial differential equations and hence we try to recover the initial vascularization solving the inverse problem. We show that the estimated vasculature could be used to efficiently predict the future tumor growth. In the second part of this thesis, we introduce a class of models dedicated to glioma, adapted both to low grade and multiform glioblastoma. In order to take into account their specificities, we include mainly two effects in the model : on the one hand, the infiltrate behaviors of gliomas, and on the other hand, the impact of brain heterogeneity, of brain anisotropy and of brain geometry on the tumor growth. Our models allow us to evaluate the efficiency of anti-angiogenic drugs and to compare it with the effect of drugs inhibiting the invasion ability of glioma. The models have been implemented in 2D and 3D in actual geometry provided by an atlas.