[ES] Las enfermedades cardiovasculares son una de las principales causas de mortalidad y morbilidad en el mundo. Entre las arritmias más comunes en adultos destaca la Fibrilación Auricular (FA), presentando una tendencia de crecimiento muy significativa, sobre todo en población de edad avanzada o con trastornos de obesidad. En el otro extremo, nos encontramos con la Miocardiopatía Arritmogénica (MCA), considerada una enfermedad rara con una prevalencia de 1:2000-5000 pero con gran afectación entre familiares directos, causante de muerte súbita cardiaca (MSC), y con un diagnóstico clínico complicado. Más allá de la FA o la MCA, existe una amplia variedad de patologías derivadas de una disfunción en la activación y conducción eléctrica del corazón. Para todas ellas, el electrocardiograma (ECG) continúa figurando como la primera y principal técnica de diagnóstico clínico, siendo una herramienta fundamental de cribado y detección de patologías relativamente económica y ampliamente accesible. Sin embargo, el diagnóstico preciso a partir de la interpretación del ECG requiere de médicos experimentados, siendo ésta una tarea que consume recursos, tiempo y que además está sujeta a la variabilidad entre observadores. Respecto a las afecciones cardiacas más comunes, conseguir un diagnóstico de forma automática que sea fiable, utilizando tanto 12 como un número reducido o único de derivaciones, sigue presentándose como un desafío. Este aspecto cobra especial relevancia con el uso cada vez más extendido de dispositivos portátiles o wearables, los cuales están ganando un gran interés para la detección temprana y preventiva de enfermedades cardiacas, registrando normalmente un número reducido de derivaciones ECG. Dicho uso masivo les confiere un gran potencial para facilitar el cribado y seguimiento de distintas afecciones en una amplia variedad de escenarios, a pesar de registrar señales de peor calidad en comparación con equipos certificados para uso clínico. El principal reto con estos dispositivos es encontrar un equilibrio adecuado entre la sensibilidad y la especificidad en la detección de ritmos cardiacos susceptibles de ser patológicos. En consecuencia, es indispensable diseñar e implementar algoritmos precisos adecuados para dispositivos móviles o portátiles capaces de detectar distintas afecciones cardiacas en registros de ECG. Respecto las afecciones cardiacas menos comunes como el caso de la MCA, es necesario incrementar la sensibilidad en la detección durante los cribados intra-familiares realizados tras una MSC. Para ello, sería posible explorar biomarcadores propios a esta enfermedad obtenidos mediante técnicas de procesado de señales ECG, además de modelos de clasificación que hagan uso de ellos, contribuyendo así a reducir el número de casos de muerte súbita. En base a lo descrito anteriormente, la presente tesis estudia las posibilidades de diagnóstico basadas en técnicas de aprendizaje y clasificación automática en dos escenarios principales. El primero aborda la detección de la FA, así como un amplio abanico de otras patologías cardiacas comunes, donde proponemos y validamos distintos modelos de clasificación de bajo consumo computacional. Todo esto, utilizando extensas bases de datos de acceso abierto, y haciendo énfasis en enfoques de derivación única, ya que son los más utilizados en dispositivos móviles e inteligentes. El segundo escenario se centra en la detección de MCA mediante las 12 derivaciones estándar del ECG, donde proponemos y validamos nuevos biomarcadores y modelos de clasificación que tratan de incrementar la sensibilidad de los cribados intra-familiares realizados tras una MSC. Para ello, utilizamos una base de datos específica de la Unidad de Cardiopatías Familiares del Hospital Universitario y Politécnico La Fe de València. / [CA] Les malalties cardiovasculars són una de les principals causes de mortalitat i morbiditat en el món. Entre les arrítmies més comunes en adults destaca la Fibril·lació Auricular (FA), presentant una tendència de creixement molt significativa, sobretot en població d'edat avançada o amb trastorns d'obesitat. En l'altre extrem, ens trobem amb la Miocardiopatia Arritmogènica (MCA), considerada una malaltia rara amb una prevalença de 1:2000-5000 però amb gran afectació entre familiars directes, causant de mort sobtada cardíaca (MSC), i amb un diagnòstic clínic complicat. Més enllà de la FA o la MCA, existeix una àmplia varietat de patologies derivades d'una disfunció en l'activació i conducció elèctrica del cor. Per a totes elles, l'electrocardiograma (ECG) continua figurant com la primera i principal tècnica de diagnòstic clínic, sent una eina fonamental de cribratge i detecció de patologies relativament econòmica i àmpliament accessible. No obstant això, el diagnòstic precís a partir de la interpretació del ECG requereix de metges experimentats, sent aquesta una tasca que consumeix recursos, temps i que a més està subjecta a la variabilitat entre observadors. Respecte a les afeccions cardíaques més comunes, aconseguir un diagnòstic de manera automàtica que siga fiable, utilitzant tant 12 com un número reduït o únic de derivacions, continua presentant-se com un desafiament. Aquest aspecte cobra especial rellevància amb l'ús cada vegada més estés de dispositius portàtils o wearables, els quals estan guanyant un gran interés per a la detecció precoç i preventiva de malalties cardíaques, registrant normalment un nombre reduït de derivacions ECG. Aquest ús massiu els confereix un gran potencial per a facilitar el cribratge i seguiment de diferents afeccions en una àmplia varietat d'escenaris, malgrat registrar senyals de pitjor qualitat en comparació amb equips certificats per a ús clínic. El principal repte amb aquests dispositius és trobar un equilibri adequat entre la sensibilitat i l'especificitat en la detecció de ritmes cardíacs susceptibles de ser patològics. En conseqüència, és indispensable dissenyar i implementar algorismes precisos adequats per a dispositius mòbils o portàtils capaços de detectar diferents afeccions cardíaques en registres de ECG. Respecte les afeccions cardíaques menys comunes com el cas de la MCA, és necessari incrementar la sensibilitat en la detecció durant els cribratges intra-familiars realitzats després d'una MSC. Per a això, seria possible explorar biomarcadors propis a aquesta malaltia obtinguts mitjançant tècniques de processament de senyals ECG, a més de models de classificació que facen ús d'ells, contribuint així a reduir el nombre de casos de mort sobtada. Sobre la base del descrit anteriorment, la present tesi estudia les possibilitats de diagnòstic basades en tècniques d'aprenentatge i classificació automàtica en dos escenaris principals. El primer aborda la detecció de la FA, així com un ampli ventall d'altres patologies cardíaques comunes, on proposem i validem diferents models de classificació de baix consum computacional. Tot això, utilitzant extenses bases de dades d'accés obert, i fent èmfasi en enfocaments de derivació única, ja que són els més utilitzats en dispositius mòbils i intel·ligents. El segon escenari se centra en la detecció de MCA mitjançant les 12 derivacions estàndard de l'ECG, on proposem i validem nous biomarcadors i models de classificació que tracten d'incrementar la sensibilitat dels cribratges intra-familiars realitzats després d'una MSC. Per a això, utilitzem una base de dades específica de la Unitat de Cardiopaties Familiars de l'Hospital Universitari i Politècnic La Fe de València. / [EN] Cardiovascular diseases are one of the leading causes of mortality and morbidity worldwide. Atrial Fibrillation (AF) stands out among adults' most common arrhythmias, presenting a very significant growth trend, especially in the elderly population or those with obesity disorders. At the other extreme, we find Arrhythmogenic Cardiomyopathy (ACM), a rare disease with a prevalence of 1:2000-5000 but great affectation among direct relatives, causing sudden cardiac death (SCD), and with a complicated clinical diagnosis. Beyond AF or ACM, there is a wide variety of pathologies derived from dysfunctions in the activation or electrical conduction of the heart. For all of them, the electrocardiogram (ECG) continues to appear as the first and foremost clinical diagnostic technique, being a fundamental tool for screening and detecting pathologies that is relatively cheap and widely accessible. However, accurate diagnosis based on ECG interpretation requires experienced physicians, as this task consumes resources, time and is subject to variability between observers. For the most common cardiac conditions, achieving a reliable diagnosis automatically, using either 12 or a smaller or single number of leads, remains a challenge. This aspect is especially relevant with the increasingly widespread use of portable or wearable devices, which are gaining significant interest for the early and preventive detection of heart disease, typically recording a reduced number of ECG leads. Such massive use gives them great potential to facilitate screening and monitoring different conditions in different scenarios, despite registering signals of lower quality compared to equipment certified for clinical use. The main challenge with these devices is finding the right balance between sensitivity and specificity in detecting pathologic heart rhythms. Consequently, designing and implementing accurate algorithms suitable for mobile or portable devices capable of detecting different cardiac conditions in ECG recordings is essential. Concerning less common cardiac conditions such as the case of ACM, it is necessary to increase the sensitivity in detection during intra-family screenings carried out after an SCD. Hence, it would be possible to explore specific biomarkers to this disease obtained through ECG signal processing techniques, as well as classification models that use them, thus contributing to reduce the number of cases of sudden death. Based on the previously described, this thesis studies the diagnostic possibilities based on machine learning and classification techniques in two main scenarios. The first deals with detecting AF and a wide range of other common cardiac pathologies, where we propose and validate different classification models with low computational consumption. All this, using extensive open access databases, and emphasizing single-lead approaches, since they are the most used in mobile and smart devices. The second scenario focuses on detecting ACM using the standard 12-lead ECG, where we propose and validate new biomarkers and classification models that try to increase the sensitivity of intra-family screenings carried out after an MSC. For this task, we used a specific database of the Familial Cardiopathies Unit of the Hospital Universitario y Politécnico La Fe de València. / Jiménez Serrano, S. (2023). Desarrollo de nuevos marcadores y clasificadores de bajo coste computacional para identificar afecciones cardiacas en registros ECG [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/196826

Ο στόχος αυτής της εργασίας ήταν η αναγνώριση έξι βασικών κινήσεων του χεριού με χρήση δύο συστημάτων. Όντας θέμα διεπιστημονικού επιπέδου έγινε μελέτη της ανατομίας των μυών του πήχη, των βιοσημάτων, της μεθόδου της ηλεκτρομυογραφίας (ΗΜΓ) και μεθόδων αναγνώρισης προτύπων. Παράλληλα, το σήμα περιείχε αρκετό θόρυβο και έπρεπε να αναλυθεί, με χρήση του EMD, να εξαχθούν χαρακτηριστικά αλλά και να μειωθεί η διαστασιμότητά τους, με χρήση των RELIEF και PCA, για βελτίωση του ποσοστού επιτυχίας ταξινόμησης. Στο πρώτο μέρος γίνεται χρήση συστήματος ΗΜΓ της Delsys αρχικά σε ένα άτομο και στη συνέχεια σε έξι άτομα με το κατά μέσο όρο επιτυχημένης ταξινόμησης, για τις έξι αυτές κινήσεις, να αγγίζει ποσοστά άνω του 80%. Το δεύτερο μέρος περιλαμβάνει την κατασκευή αυτόνομου συστήματος ΗΜΓ με χρήση του Arduino μικροελεγκτή, αισθητήρων ΗΜΓ και ηλεκτροδίων, τα οποία είναι τοποθετημένα σε ένα ελαστικό γάντι. Τα αποτελέσματα ταξινόμησης σε αυτή την περίπτωση αγγίζουν το 75%. / The aim of this work was to identify six basic movements of the hand using two systems. Being an interdisciplinary topic, there has been conducted studying in the anatomy of forearm muscles, biosignals, the method of electromyography (EMG) and methods of pattern recognition. Moreover, the signal contained enough noise and had to be analyzed, using EMD, to extract features and to reduce its dimensionality, using RELIEF and PCA, to improve the success rate of classification. The first part uses an EMG system of Delsys initially for an individual and then for six people with the average successful classification, for these six movements at rates of over 80%. The second part involves the construction of an autonomous system EMG using an Arduino microcontroller, EMG sensors and electrodes, which are arranged in an elastic glove. Classification results in this case reached 75% of success.

Given satellite images with accompanying pixel classifications and elevation data, we propose different solutions to object detection. The first method uses hierarchical clustering for segmentation and then employs different methods of classification. One of these classification methods used domain knowledge to classify objects while the other used Support Vector Machines. Additionally, a combination of three Support Vector Machines were used in a hierarchical structure which out-performed the regular Support Vector Machine method in most of the evaluation metrics. The second approach is more conventional with different types of Convolutional Neural Networks. A segmentation network was used as well as a few detection networks and different fusions between these. The Convolutional Neural Network approach proved to be the better of the two in terms of precision and recall but the clustering approach was not far behind. This work was done using a relatively small amount of data which potentially could have impacted the results of the Machine Learning models in a negative way.

Este trabalho apresenta um sistema computacional que utiliza técnicas de Aprendizado de Máquina (AM) para auxiliar o diagnóstico oftalmológico. Trata-se de um sistema de medidas objetivas e automáticas dos principais vícios de refração ocular, astigmatismo, hipermetropia e miopia. O sistema funcional desenvolvido aplica técnicas convencionais de processamento a imagens do olho humano fornecidas por uma técnica de aquisição chamada Hartmann-Shack (HS), ou Shack-Hartmann (SH), com o objetivo de extrair e enquadrar a região de interesse e remover ruídos. Em seguida, vetores de características são extraídos dessas imagens pela técnica de transformada wavelet de Gabor e, posteriormente, analisados por técnicas de AM para diagnosticar os possíveis vícios refrativos presentes no globo ocular representado. Os resultados obtidos indicam a potencialidade dessa abordagem para a interpretação de imagens de HS de forma que, futuramente, outros problemas oculares possam ser detectados e medidos a partir dessas imagens. Além da implementação de uma nova abordagem para a medição dos vícios refrativos e da introdução de técnicas de AM na análise de imagens oftalmológicas, o trabalho contribui para a investigação da utilização de Máquinas de Vetores Suporte e Redes Neurais Artificiais em sistemas de Entendimento/Interpretação de Imagens (Image Understanding). O desenvolvimento deste sistema permite verificar criticamente a adequação e limitações dessas técnicas para a execução de tarefas no campo do Entendimento/Interpretação de Imagens em problemas reais. / This work presents a computational system that uses Machine Learning (ML) techniques to assist in ophthalmological diagnosis. The system developed produces objective and automatic measures of ocular refraction errors, namely astigmatism, hypermetropia and myopia from functional images of the human eye acquired with a technique known as Hartmann-Shack (HS), or Shack-Hartmann (SH). Image processing techniques are applied to these images in order to remove noise and extract the regions of interest. The Gabor wavelet transform technique is applied to extract feature vectors from the images, which are then input to ML techniques that output a diagnosis of the refractive errors in the imaged eye globe. Results indicate that the proposed approach creates interesting possibilities for the interpretation of HS images, so that in the future other types of ocular diseases may be detected and measured from the same images. In addition to implementing a novel approach for measuring ocular refraction errors and introducing ML techniques for analyzing ophthalmological images, this work investigates the use of Artificial Neural Networks and Support Vector Machines (SVMs) for tasks in Image Understanding. The description of the process adopted for developing this system can help in critically verifying the suitability and limitations of such techniques for solving Image Understanding tasks in \"real world\" problems.

L'analyse automatique de différences anatomiques en neuroimagerie a de nombreuses applications pour la compréhension et l'aide au diagnostic de pathologies neurologiques. Récemment, il y a eu un intérêt croissant pour les méthodes de classification telles que les machines à vecteurs supports pour dépasser les limites des méthodes univariées traditionnelles. Cette thèse a pour thème l'apprentissage automatique pour l'analyse de populations et la classification de patients en neuroimagerie. Nous avons tout d'abord comparé les performances de différentes stratégies de classification, dans le cadre de la maladie d'Alzheimer à partir d'images IRM anatomiques de 509 sujets de la base de données ADNI. Ces différentes stratégies prennent insuffisamment en compte la distribution spatiale des \textit{features}. C'est pourquoi nous proposons un cadre original de régularisation spatiale et anatomique des machines à vecteurs supports pour des données de neuroimagerie volumiques ou surfaciques, dans le formalisme de la régularisation laplacienne. Cette méthode a été appliquée à deux problématiques cliniques: la maladie d'Alzheimer et les accidents vasculaires cérébraux. L'évaluation montre que la méthode permet d'obtenir des résultats cohérents anatomiquement et donc plus facilement interprétables, tout en maintenant des taux de classification élevés. / Brain image analyses have widely relied on univariate voxel-wise methods. In such analyses, brain images are first spatially registered to a common stereotaxic space, and then mass univariate statistical tests are performed in each voxel to detect significant group differences. However, the sensitivity of theses approaches is limited when the differences involve a combination of different brain structures. Recently, there has been a growing interest in support vector machines methods to overcome the limits of these analyses.This thesis focuses on machine learning methods for population analysis and patient classification in neuroimaging. We first evaluated the performances of different classification strategies for the identification of patients with Alzheimer's disease based on T1-weighted MRI of 509 subjects from the ADNI database. However, these methods do not take full advantage of the spatial distribution of the features. As a consequence, the optimal margin hyperplane is often scattered and lacks spatial coherence, making its anatomical interpretation difficult. Therefore, we introduced a framework to spatially regularize support vector machines for brain image analysis based on Laplacian regularization operators. The proposed framework was then applied to the analysis of stroke and of Alzheimer's disease. The results demonstrated that the proposed classifier generates less-noisy and consequently more interpretable feature maps with no loss of classification performance.

Les avancées technologiques ont permis le stockage de grandes masses de données en termes de taille (nombre d’observations) et de dimensions (nombre de variables).Ces données nécessitent de nouvelles méthodes, notamment en modélisation prédictive (data science ou science des données), de traitement statistique adaptées à leurs caractéristiques. Dans le cadre de cette thèse, nous nous intéressons plus particulièrement aux données dont le nombre de variables est élevé comparé au nombre d’observations.Pour ces données, une réduction du nombre de variables initiales, donc de dimensions, par la sélection d’un sous-ensemble optimal, s’avère nécessaire, voire indispensable.Elle permet de réduire la complexité, de comprendre la structure des données et d’améliorer l’interprétation des résultats et les performances du modèle de prédiction ou de classement en éliminant les variables bruit et/ou redondantes.Nous nous intéressons plus précisément à la sélection de variables dans le cadre de l’apprentissage supervisé et plus spécifiquement de la discrimination à catégories multiples dite multi-classes. L’objectif est de proposer de nouvelles méthodes de sélection de variables pour les modèles de discrimination multi-classes appelés Machines à Vecteurs Support Multiclasses (MSVM).Deux approches sont proposées dans ce travail. La première, présentée dans un contexte classique, consiste à sélectionner le sous-ensemble optimal de variables en utilisant le critère de "la borne rayon marge" majorante du risque de généralisation des MSVM. Quant à la deuxième approche, elle s’inscrit dans un contexte topologique et utilise la notion de graphes de voisinage et le critère de degré d’équivalence topologique en discrimination pour identifier les variables pertinentes qui constituent le sous-ensemble optimal du modèle MSVM.L’évaluation de ces deux approches sur des données simulées et d’autres réelles montre qu’elles permettent de sélectionner, à partir d’un grand nombre de variables initiales, un nombre réduit de variables explicatives avec des performances similaires ou encore meilleures que celles obtenues par des méthodes concurrentes. / The technological progress has allowed the storage of large amounts of data in terms of size (number of observations) and dimensions (number of variables). These data require new methods, especially for predictive modeling (data science), of statistical processing adapted to their characteristics. In this thesis, we are particularly interested in the data with large numberof variables compared to the number of observations.For these data, reducing the number of initial variables, hence dimensions, by selecting an optimal subset is necessary, even imperative. It reduces the complexity, helps to understand the data structure, improves the interpretation of the results and especially enhances the performance of the forecasting model by eliminating redundant and / or noise variables.More precisely, we are interested in the selection of variables in the context of supervised learning, specifically of multiclass discrimination. The objective is to propose some new methods of variable selection for multiclass discriminant models called Multiclass Support Vector Machines (MSVM).Two approaches are proposed in this work. The first one, presented in a classical context, consist in selecting the optimal subset of variables using the radius margin upper bound of the generalization error of MSVM. The second one, proposed in a topological context, uses the concepts of neighborhood graphs and the degree of topological equivalence in discriminationto identify the relevant variables and to select the optimal subset for an MSVM model.The evaluation of these two approaches on simulated and real data shows that they can select from a large number of initial variables, a reduced number providing equal or better performance than those obtained by competing methods.

The gains made by current deep-learning techniques have often come with the price tag of big data and where that data is not available, a new solution must be found. Such is the case for accented and noisy speech where large databases do not exist and data augmentation techniques, which are less than perfect, present an even larger obstacle. Another problem is that state-of-the-art results are rarely reproducible because they use proprietary datasets, pretrained networks and/or weight initializations from other larger networks. An example of a low resource scenario exists even in the fifth largest land in the world; home to most of the speakers of the seventh most spoken language on earth. Brazil is the leader in the Latin-American economy and as a BRIC country aspires to become an ever-stronger player in the global marketplace. Still, English proficiency is low, even for professionals in businesses and universities. Low intelligibility and strong accents can damage professional credibility. It has been established in the literature for foreign language teaching that it is important that adult learners are made aware of their errors as outlined by the Noticing Theory, explaining that a learner is more successful when he is able to learn from his own mistakes. An essential objective of this dissertation is to classify phonemes in the acoustic model which is needed to properly identify phonemic errors automatically. A common belief in the community is that deep learning requires large datasets to be effective. This happens because brute force methods create a highly complex hypothesis space which requires large and complex networks which in turn demand a great amount of data samples in order to generate useful networks. Besides that, the loss functions used in neural learning does not provide statistical learning guarantees and only guarantees the network can memorize the training space well. In the case of accented or noisy speech where a new sample can carry a great deal of variation from the training samples, the generalization of such models suffers. The main objective of this dissertation is to investigate how more robust acoustic generalizations can be made, even with little data and noisy accented-speech data. The approach here is to take advantage of raw feature extraction provided by deep learning techniques and instead focus on how learning guarantees can be provided for small datasets to produce robust results for acoustic modeling without the dependency of big data. This has been done by careful and intelligent parameter and architecture selection within the framework of the statistical learning theory. Here, an intelligently defined CNN architecture, together with context windows and a knowledge-driven hierarchical tree of SVM classifiers achieves nearly state-of-the-art frame-wise phoneme recognition results with absolutely no pretraining or external weight initialization. A goal of this thesis is to produce transparent and reproducible architectures with high frame-level accuracy, comparable to the state of the art. Additionally, a convergence analysis based on the learning guarantees of the statistical learning theory is performed in order to evidence the generalization capacity of the model. The model achieves 39.7% error in framewise classification and a 43.5% phone error rate using deep feature extraction and SVM classification even with little data (less than 7 hours). These results are comparable to studies which use well over ten times that amount of data. Beyond the intrinsic evaluation, the model also achieves an accuracy of 88% in the identification of epenthesis, the error which is most difficult for Brazilian speakers of English This is a 69% relative percentage gain over the previous values in the literature. The results are significant because it shows how deep feature extraction can be applied to little data scenarios, contrary to popular belief. The extrinsic, task-based results also show how this approach could be useful in tasks like automatic error diagnosis. Another contribution is the publication of a number of freely available resources which previously did not exist, meant to aid future researches in dataset creation. / Os ganhos obtidos pelas atuais técnicas de aprendizado profundo frequentemente vêm com o preço do big data e nas pesquisas em que esses grandes volumes de dados não estão disponíveis, uma nova solução deve ser encontrada. Esse é o caso do discurso marcado e com forte pronúncia, para o qual não existem grandes bases de dados; o uso de técnicas de aumento de dados (data augmentation), que não são perfeitas, apresentam um obstáculo ainda maior. Outro problema encontrado é que os resultados do estado da arte raramente são reprodutíveis porque os métodos usam conjuntos de dados proprietários, redes prétreinadas e/ou inicializações de peso de outras redes maiores. Um exemplo de um cenário de poucos recursos existe mesmo no quinto maior país do mundo em território; lar da maioria dos falantes da sétima língua mais falada do planeta. O Brasil é o líder na economia latino-americana e, como um país do BRIC, deseja se tornar um participante cada vez mais forte no mercado global. Ainda assim, a proficiência em inglês é baixa, mesmo para profissionais em empresas e universidades. Baixa inteligibilidade e forte pronúncia podem prejudicar a credibilidade profissional. É aceito na literatura para ensino de línguas estrangeiras que é importante que os alunos adultos sejam informados de seus erros, conforme descrito pela Noticing Theory, que explica que um aluno é mais bem sucedido quando ele é capaz de aprender com seus próprios erros. Um objetivo essencial desta tese é classificar os fonemas do modelo acústico, que é necessário para identificar automaticamente e adequadamente os erros de fonemas. Uma crença comum na comunidade é que o aprendizado profundo requer grandes conjuntos de dados para ser efetivo. Isso acontece porque os métodos de força bruta criam um espaço de hipóteses altamente complexo que requer redes grandes e complexas que, por sua vez, exigem uma grande quantidade de amostras de dados para gerar boas redes. Além disso, as funções de perda usadas no aprendizado neural não fornecem garantias estatísticas de aprendizado e apenas garantem que a rede possa memorizar bem o espaço de treinamento. No caso de fala marcada ou com forte pronúncia, em que uma nova amostra pode ter uma grande variação comparada com as amostras de treinamento, a generalização em tais modelos é prejudicada. O principal objetivo desta tese é investigar como generalizações acústicas mais robustas podem ser obtidas, mesmo com poucos dados e/ou dados ruidosos de fala marcada ou com forte pronúncia. A abordagem utilizada nesta tese visa tirar vantagem da raw feature extraction fornecida por técnicas de aprendizado profundo e obter garantias de aprendizado para conjuntos de dados pequenos para produzir resultados robustos para a modelagem acústica, sem a necessidade de big data. Isso foi feito por meio de seleção cuidadosa e inteligente de parâmetros e arquitetura no âmbito da Teoria do Aprendizado Estatístico. Nesta tese, uma arquitetura baseada em Redes Neurais Convolucionais (RNC) definida de forma inteligente, junto com janelas de contexto e uma árvore hierárquica orientada por conhecimento de classificadores que usam Máquinas de Vetores Suporte (Support Vector Machines - SVMs) obtém resultados de reconhecimento de fonemas baseados em frames quase no estado da arte sem absolutamente nenhum pré-treinamento ou inicialização de pesos de redes externas. Um objetivo desta tese é produzir arquiteturas transparentes e reprodutíveis com alta precisão em nível de frames, comparável ao estado da arte. Adicionalmente, uma análise de convergência baseada nas garantias de aprendizado da teoria de aprendizagem estatística é realizada para evidenciar a capacidade de generalização do modelo. O modelo possui um erro de 39,7% na classificação baseada em frames e uma taxa de erro de fonemas de 43,5% usando raw feature extraction e classificação com SVMs mesmo com poucos dados (menos de 7 horas). Esses resultados são comparáveis aos estudos que usam bem mais de dez vezes essa quantidade de dados. Além da avaliação intrínseca, o modelo também alcança uma precisão de 88% na identificação de epêntese, o erro que é mais difícil para brasileiros falantes de inglês. Este é um ganho relativo de 69% em relação aos valores anteriores da literatura. Os resultados são significativos porque mostram como raw feature extraction pode ser aplicada a cenários de poucos dados, ao contrário da crença popular. Os resultados extrínsecos também mostram como essa abordagem pode ser útil em tarefas como o diagnóstico automático de erros. Outra contribuição é a publicação de uma série de recursos livremente disponíveis que anteriormente não existiam, destinados a auxiliar futuras pesquisas na criação de conjuntos de dados.

No description available.

Dans ce travail de thèse, nous présentons nos travaux qui portent sur l’identification des cibles en général par un radar Ultra-Large Bande (ULB) et en particulier l’identification des cibles dont la surface équivalente radar est faible telles que les piétons et les cyclistes. Ce travail se décompose en deux parties principales, la détection et la reconnaissance. Dans la première approche du processus de détection, nous avons proposé et étudié un détecteur de radar ULB robuste qui fonctionne avec des données radar 1-D (A-scan) à une dimension. Il exploite la combinaison des statistiques d’ordres supérieurs et du détecteur de seuil automatique connu sous le nom de CA-CFAR pour Cell-Averaging Constant False Alarm Rate. Cette combinaison est effectuée en appliquant d’abord le HOS sur le signal reçu afin de supprimer une grande partie du bruit. Puis, après avoir éliminé le bruit du signal radar reçu, nous implémentons le détecteur de seuil automatique CA-CFAR. Ainsi, cette combinaison permet de disposer d’un détecteur de radar ULB à seuil automatique robuste. Afin d’améliorer le taux de détection et aller plus loin dans le traitement, nous avons évalué l’approche des données radar 2-D (B-Scan) à deux dimensions. Dans un premier temps, nous avons proposé une nouvelle méthode de suppression du bruit, qui fonctionne sur des données B-Scan. Il s’agit d’une combinaison de WSD et de HOS. Pour évaluer les performances de cette méthode, nous avons fait une étude comparative avec d’autres techniques de suppression du bruit telles que l’analyse en composantes principales, la décomposition en valeurs singulières, la WSD, et la HOS. Les rapports signal à bruit -SNR- des résultats finaux montrent que les performances de la combinaison WSD et HOS sont meilleures que celles des autres méthodes rencontrées dans la littérature. A la phase de reconnaissance, nous avons exploité les données des deux approches à 1-D et à 2-D obtenues à partir du procédé de détection. Dans la première approche à 1-D, les techniques SVM et le DBN sont utilisées et évaluées pour identifier la cible en se basant sur la signature radar. Les résultats obtenus montrent que la technique SVM donne de bonnes performances pour le système proposé où le taux de reconnaissance global moyen atteint 96,24%, soit respectivement 96,23%, 95,25% et 97,23% pour le cycliste, le piéton et la voiture. Dans la seconde approche à 1-D, les performances de différents types d’architectures DBN composées de différentes couches ont été évaluées et comparées. Nous avons constaté que l’architecture du réseau DBN avec quatre couches cachées est meilleure et la précision totale moyenne peut atteindre 97,80%. Ce résultat montre que les performances obtenues avec le DBN sont meilleures que celles obtenues avec le SVM (96,24%) pour ce système de reconnaissance de cible utilisant un radar ULB. Dans l’approche bidimensionnelle, le réseau de neurones convolutifs a été utilisé et évalué. Nous avons proposé trois architectures de CNN. La première est le modèle modifié d’Alexnet, la seconde est une architecture avec les couches de convolution arborescentes et une couche entièrement connectée, et la troisième est une architecture avec les cinq couches de convolution et deux couches entièrement connectées. Après comparaison et évaluation des performances de ces trois architectures proposées nous avons constaté que la troisième architecture offre de bonnes performances par rapport aux autres propositions avec une précision totale moyenne qui peut atteindre 99,59%. Enfin, nous avons effectué une étude comparative des performances obtenues avec le CNN, DBN et SVM. Les résultats montrent que CNN a les meilleures performances en termes de précision par rapport à DBN et SVM. Cela signifie que l’utilisation de CNN dans les données radar bidimensionnels permet de classer correctement les cibles radar ULB notamment pour les cibles à faible SER et SNR telles que les cyclistes ou les piétons. / In this thesis work, we focused on the study and development of a system identification using UWB-Ultra-Wide-Band short range radar to detect the objects and particularly the vulnerable road users (VRUs) that have low RCS-Radar Cross Section- such as cyclist and pedestrian. This work is composed of two stages i.e. detection and recognition. In the first approach of detection stage, we have proposed and studied a robust UWB radar detector that works on one dimension 1-D radar data ( A-scan). It relies on a combination of Higher Order Statistics (HOS) and the well-known CA-CFAR (Cell-Averaging Constant False Alarm Rate) detector. This combination is performed by firstly applying the HOS to the received radar signal in order to suppress the noise. After eliminating the noise of the received radar signal, we apply the CA-CFAR detector. By doing this combination, we finally have an UWB radar detector which is robust against the noise and works with the adaptive threshold. In order to enhance the detection performance, we have evaluated the approach of using two dimensions 2-D (B-Scan) radar data. In this 2-D radar approach, we proposed a new method of noise suppression, which works on this B-Scan data. The proposed method is a combination of WSD (Wavelet Shrinkage Denoising) and HOS. To evaluate the performance of this method, we performed a comparative study with the other noise removal methods in literature including Principal Component Analysis (PCA), Singular Value Decomposition (SVD), WSD and HOS. The Signal-to-Noise Ratio (SNR) of the final result has been computed to compare the effectiveness of individual noise removal techniques. It is observed that a combination of WSD and HOS has better capability to remove the noise compared to that of the other applied techniques in the literature; especially it is found that it allows to distinguish efficiency the pedestrian and cyclist over the noise and clutters whereas other techniques are not showing significant result. In the recognition phase, we have exploited the data from the two approaches 1-D and 2-D, obtained from the detection method. In the first 1-D approach, Support Vector Machines (SVM) and Deep Belief Networks (DBN) have been used and evaluated to identify the target based on the radar signature. The results show that the SVM gives good performances for the proposed system where the total recognition accuracy rate could achieve up to 96,24%. In the second approach of this 1-D radar data, the performance of several DBN architectures compose of different layers have been evaluated and compared. We realised that the DBN architecture with four hidden layers performs better than those of with two or three hidden layers. The results show also that this architecture achieves up to 97.80% of accuracy. This result also proves that the performance of DBN is better than that of SVM (96.24%) in the case of UWB radar target recognition system using 1-D radar signature. In the 2-D approach, the Convolutional Neural Network (CNN) has been exploited and evaluated. In this work, we have proposed and investigated three CNN architectures. The first architecture is the modified of Alexnet model, the second is an architecture with three convolutional layers and one fully connected layer, and the third is an architecture with five convolutional layers and two fully connected layers. The performance of these proposed architectures have been evaluated and compared. We found that the third architecture has a good performance where it achieves up to 99.59% of accuracy. Finally, we compared the performances obtained using CNN, DBN and SVM. The results show that CNN gives a better result in terms of accuracy compared to that of DBN and SVM. It allows to classify correctly the UWB radar targets like cyclist and pedestrian.

Cette thèse s'inscrit dans le cadre des études menées sur les effets biologiques et sanitaires des radiofréquences associées à la téléphonie mobile. Le but de ce travail est d'évaluer l'influence des radiofréquences émises par le téléphone mobile sur l'activité corticale chez l'homme. Tout d'abord, un état des connaissances scientifiques relatives aux effets biologiques et sanitaires des radiofréquences associées à la téléphonie mobile a été dressé. Ensuite, on effectue une description du matériel et des méthodes d'acquisition des potentiels évoqués auditifs (PEA) caractéristiques de l'activité corticale auditive. De plus, on a sélectionné les électrodes qui recueillent les signaux les plus caractéristiques de PEA. Des variables dont certaines caractéristiques électrophysiologiques des PEA sont extraites et font l'objet d'une analyse statistique. Une sélection, basée sur la classification par Support Vector Machines, met en exergue les variables plus perturbées par les radiofréquences. Enfin, l'analyse statistique par électrode permet de localiser les effets des radiofréquences. On se consacre enfin à l'étude des signaux électroencéphalographiques en étudiant les coefficients de corrélations entre spectres et les densités spectrales de puissance par bande de fréquences. En conclusion, les radiofréquences émises par le téléphone mobile semblent montrer une incidence sur les potentiels évoqués auditifs et les signaux électroencéphalographiques. Cependant, aucun apport de causalité entre ces effets biologiques sur l'activité corticale et les effets néfastes sur la santé humaine ne peut être démontré à ce jour.