La technique de haute intensité ultrasons focalisés (HIFU) est maintenant largement utilisée pour le traitement du cancer, grâce à son avantage non-invasif. Dans un système de HIFU, une matrice de transducteurs à ultrasons est pilotée en phase pour produire un faisceau focalisé d'ultrasons (1M ~ 10 MHz) dans une petite zone de l'emplacement de la cible sur le cancer dans le corps. La plupart des systèmes HIFU sont guidées par imagerie par résonance magnétique (IRM) dans de nos jours. Dans cette étude de doctorat, un amplificateur de puissance de classe D en demi-pont et un système d'accord automatique d'impédance sont proposés. Tous deux circuits proposés sont compatibles avec le système IRM. L'amplificateur de puissance proposé a été réalisé par un circuit imprimé (PCB) avec des composants discrets. Selon les résultats du test, il a rendement de conversion en puissance de 82% pour une puissance de sortie conçue de 1,25W à une fréquence de travail de 3MHz. Le système d'accord automatique d'impédance proposé a été conçu en deux versions: une version en PCB et une version en circuit intégré (IC). Contrairement aux systèmes d'accord automatique proposés dans la littérature, il n'y a pas besoin de l'unité de microcontrôleur (MCU) ou de l'ordinateur dans la conception proposée. D'ailleurs, sans l'aide de composants magnétiques volumineux, ce système d'auto-réglage est entièrement compatible avec l'équipement IRM. La version en PCB a été conçue pour vérifier le principe du système proposé, et il est également utilisé pour guider à la conception du circuit intégré. La réalisation en PCB occupe une surface de 110cm². Les résultats des tests ont confirmé la performance attendue. Le système d'auto-tuning proposé peut parfaitement annuler l'impédance imaginaire du transducteur, et il peut également compenser l'impédance de la dérive causée par les variations inévitables (variation de température, dispersion technique, etc.). La conception du système d'auto-réglage en circuit intégré a été réalisé avec une technologie CMOS (C35B4C3) fournies par Austrian Micro Systems (AMS). La surface occupée par le circuit intégré est seulement de 0,42mm². Le circuit intégré conçu est capable de fonctionner à une large gamme de fréquence tout en conservant une consommation d'énergie très faible (137 mW). D'après les résultats de la simulation, le rendement de puissance de ce circuit peut être amélioré jusqu'à 20% comparant à celui utilisant le réseau d'accord statique. / High intensity focused ultrasound (HIFU) technology is now broadly used for cancer treatment, thanks to its non-invasive property. In a HIFU system, a phased array of ultrasonic transducers is utilized to generate a focused beam of ultrasound (1M~10MHz) into a small area of the cancer target within the body. Most HIFU systems are guided by magnetic resonance imaging (MRI) in nowadays. In this PhD study, a half-bridge class D power amplifier and an automatic impedance tuning system are proposed. Both the class D power amplifier and the auto-tuning system are compatible with MRI system. The proposed power amplifier is implemented by a printed circuit board (PCB) circuit with discrete components. According to the test results, it has a power efficiency of 82% designed for an output power of 3W at 1.25 MHz working frequency. The proposed automatic impedance tuning system has been designed in two versions: a PCB version and an integrated circuit (IC) version. Unlike the typical auto-impedance tuning networks, there is no need of microprogrammed control unit (MCU) or computer in the proposed design. Besides, without using bulky magnetic components, this auto-tuning system is completely compatible with MRI equipment. The PCB version was designed to verify the principle of the proposed automatic impedance tuning system, and it is also used to help the design of the integrated circuit. The PCB realization occupies a surface of 110cm². The test results confirmed the expected performance. The proposed auto-tuning system can perfectly cancel the imaginary impedance of the transducer, and it can also compensate the impedance drifting caused by unavoidable variations (temperature variation, technical dispersion, etc.). The IC design of the auto-tuning system is realized in a CMOS process (C35B4C3) provided by Austrian Micro Systems (AMS). The die area of the integrated circuit is only 0.42mm². This circuit design can provide a wide working frequency range while keeping a very low power consumption (137 mW). According to the simulation results, the power efficiency can be improved can up to 20% by using this auto-tuning circuit compared with that using the static tuning network.

Orientador: Michael John Brennan / Abstract: The advances in microelectromechanical systems to power supply devices for monitoring the structural integrity, mainly for applications in remote areas or with difficult access, have made energy harvesting from ambient vibration a highly researched topic. Although several electromechanical mechanisms have been proposed for applications in a variety of fields, harvesting energy from railway track vibrations is relatively new, and only a small amount of research is ongoing in this topic. This thesis aims to determine the factors that govern the sleeper vertical vibration induced by a passing train, develop a methodology to emulate this vibration in a laboratory-based system and to derive a model of a load resistance attached to a linear electromagnetic transducer to determine the optimum energy harvested. The study shows the importance of knowing the vibration behaviour of the sleeper in order to tune the energy harvester to the correct excitation frequency. A comparison between piezoelectric and electromagnetic transducers is performed using a two-port network model to determine which transducer performs better for low frequency vibration. To emulate sleeper vibration due to a passing train using an electrodynamic shaker, a compensator filter is designed to remove the system dynamics. An analytical investigation into the energy dissipated by a load resistance attached to an electromagnetic transducer when subject to a time-limited base excitation, which may include the transie... (Complete abstract click electronic access below) / Resumo: Avanços em sistemas micro eletromecânico para fornecer energia à dispositivos para monitoramento de integridade estrutural, principalmente para aplicações em áreas remotas ou de difícil acesso, fez de extração de energia a partir de vibrações contidas no ambiente um tópico de pesquisa em destaque. Apesar de que diversos mecanismos eletromecânicos já foram propostos em uma variedade de aplicações, extração de energia à partir de vibrações em linhas férreas é, relativamente, novo e somente algumas pesquisas estão sendo realizadas sobre esse tópico. Esta Tese busca determinar os fatores que governam a vibração vertical do dormente induzida devido a passagem de um trem, desenvolver uma metodologia para emular esta vibração em um sistema em laboratório e derivar um modelo para um transdutor linear eletromagnético com uma carga resistiva acoplada. O estudo mostrou a importância de conhecer o comportamento vibracional do dormente de modo a sintonizar o extrator de energia com a frequência correta de excitação. Uma comparação entre transdutores piezoelétrico e eletromagnético é realizado utilizando um modelo quadripolo para determinar qual transdutor tem um melhor desempenho para vibrações em baixa frequência. Para emular a vibração do dormente devido a passagem de um trem em um agitador eletrodinâmico, um filtro compensador é projetado para remover a dinâmica do sistema. Uma investigação analítica da energia dissipada pela carga resistiva acoplada à um transdutor eletromagnético qua... (Resumo completo, clicar acesso eletrônico abaixo) / Doutor

Numerous vibrating electromechanical systems lack a rigid connection to the inertial frame. An artificial inertial frame can be generated by a shaker, which compensates for vibrations. In this article, we present an encapsulated and perforated unimorph bending plate for this purpose. Vibrations can be compensated up to the first eigenfrequency of the system. As basis for an efficient system simulation and optimization, a new three-port multi-domain network model was developed. An extension qualifies the network for the simulation of the acoustical behavior inside the capsule. Network parameters are determined using finite element simulations. The dynamic behavior of the network model agrees with the finite element simulation results up to the first resonance of the system. The network model was verified by measurements on a laboratory setup, too. Furthermore, the network model could be simplified and was applied to determine the influence of various parameters on the stabilization performance of the plate transducer. The performance of the piezoelectric bending plate for position stabilization had been in addition investigated experimentally by measurements on a macroscopic capsule.

No description available.

[pt] A velocidade da onda de pulso (VOP) tem sido identificada como o padrão-ouro para avaliação da rigidez arterial e, recentemente, vem sendo reconhecida como um importante indicador no diagnóstico e tratamento de doenças cardiovasculares. Atualmente, já existem dispositivos comerciais capazes de efetuar a medição da VOP, entretanto, ainda exigem um investimento financeiro significativo e alguns requerem um treinamento especializado para seu correto uso. Os, transdutores de pressão atuais são majoritariamente baseados em sensores piezoresistivos, piezoelétricos e capacitivos. Entretanto, pesquisas recentes demostraram que transdutores de pressão que utilizam sensores magnéticos baseados na magnetoimpedância gigante (GMI) apresentam elevada sensibilidade. Tendo em vista que a VOP é um importante indicador do risco de distúrbios cardiovasculares, e considerando os potenciais beneficios dos sensores GMI em relação às demais alternativas, esta tese de doutorado buscou utilizar-se destes elementos sensores a fim de desenvolver um sistema de medição portátil, não-invasivo, de baixo custo, acessível e simples de usar, capaz de efetuar a medição da VOP. Neste intuito, foram desenvolvidos transdutores de alta sensibilidade, baseados nas características de fase da impedância de sensores de Magnetoimpedância Gigante, destinados à medição da velocidade da onda de pulso arterial. A fim de se otimizar as características de desempenho dos transdutores, foram realizadas avaliações teórico-computacionais dos transdutores na configuração em malha aberta e fechada, bem como ensaios experimentais dos protótipos projetados. As caracterizações e ensaios experimentais realizados com o transdutor de pressão em malha aberta resultaram em uma sensibilidade de 59,6 mV/kPa, e resolução de 192,8 Pa para uma média de 30 amostras, na banda de passagem de 1000 Hz. Por outro lado, a configuração em malha fechada apresentou uma sensibilidade de 54,2 mV/kPa, e resolução de 206,0 Pa para uma média de 30 amostras, na banda de passagem de 32 Hz. Tendo em vista os valores de sensibilidade e resolução obtidos, propõe-se empregar o sistema de transdução de pressão que incorpora uma câmara incompressível para amplificação mecânica, na medição de ondas de pulso arterial. Neste protótipo, uma pequena membrana semirrígida localizada na superfície da câmara incompressível é posicionada sobre a superfície da pele, próxima à artéria de interesse. Deste modo, pequenas mudanças de pressão na superfície da pele, causadas pela onda de pulso arterial, provocam uma variação do campo magnético sobre o elemento sensor. Por outra parte, devido à alta sensibilidade apresentada pelo transdutor magnetico (magnetômetro GMI) na configuração de malha aberta (0,2 mV/nT) e de malha fechada (0,19 mV/nT), estes foram usados para medir diretamente a forma de onda do pulso arterial, sem utilizar uma câmara incompressível para transdução mecânica. Nesta medição, considerando a adequada resolução espacial para as demandas anatômicas, utiliza-se um pequeno marcador magnético, envolto por uma fita adesiva hipoalergênica e flexível, aderida á região da pele sobre a artéria de interesse, e aproxima-se o sensor magnético GMI da superfície da pele onde o marcador foi colocado. Finalmente, as configurações propostas foram analisadas e comparadas, a fim de se identificar aquela com melhor desempenho, a qual foi utilizada para medição da VOP. Como o estudo envolve o registro da onda de pulso em participantes da pesquisa, o projeto foi submetido à apreciação e aprovado pela Comissão da Câmara de Ética em Pesquisa da Pontifícia Universidade Católica do Rio de Janeiro (PUC-Rio) 045/2020 – Protocolo 83/2020. Espera-se que o dispositivo desenvolvido contribua para o avanço tecnológico do ferramental utilizado no setor da saúde. / [en] Pulse wave velocity (PWV) is considered the gold standard for assessing arterial stiffness and recently, it has been recognized as an important indicator in the diagnosis and treatment of cardiovascular disease. Currently, there are commercial devices capable of measuring PWV, however, significant investments are required and some devices requires specialized training for their correct use. Conventional pressure-sensing devices are mainly based on piezoresistive, piezoelectric and capacitive sensors. Recent investigations, however, show that pressure transducer using magnetic sensors based on the giant Magnetoimpedance (GMI) present high-sensitivity. Considering that, PWV is a significant risk factor for future cardiovascular disease and in view of some of the advantages of GMI sensors in relation to another sensing technologies, this doctoral thesis aims to develop a portable measurement system, non-invasive, low-cost, accessible and simple to use, capable of measuring PWV. For this purpose, we have developed a high-sensitivity transducers based on the impedance phase characteristics of GMI sensors, for measuring the arterial pulse wave velocity. In order to improve the performance characteristics of the transducers, computational and theoretical analysis in open and closed loop configuration were performed. The characterizations and experimental tests performed with the open-loop pressure transducer resulted in a sensitivity of 59.6 mV/kPa, and resolution of 192.8 Pa for an average of 30 samples, in the 1000 Hz passband. On the other hand, the closed-loop configuration presented a sensitivity of 54.2 mV/kPa, and a resolution of 206.0 Pa for an average of 30 samples, in the 32 Hz passband. In view of the considerable sensitivity and resolution obtained, it is proposed to employ a pressure transduction system that incorporates an incompressible chamber for mechanical amplification, in the measurement of arterial pulse waves. In this prototype, a small semi-rigid membrane located on the surface of the incompressible chamber is positioned over the surface of the skin, close to the artery of interest. In this way, small pressure changes on the skin surface, caused by the arterial pulse wave, cause a variation of the magnetic field on the sensing element. On the other hand, due to the high sensitivity presented by the magnetic transducer (GMI magnetometer) in the open-loop (0.2 mV/nT) and closed-loop (0.19 mV/nT) configurations, they were used to measure the shape pulse waveform without using an incompressible chamber for mechanical transduction. In this test, considering the adequate spatial resolution for the anatomical demands, a small magnetic marker is used, the magnetic marker is attached to the skin region over the artery of interest, and the GMI magnetic sensor is approached near the marker placed of the skin surface. Finally, the proposed configurations were analyzed and compared in order to identify the one with the best performance, which was used to measure PWV. As the study involves recording the pulse wave in research participants, the project was submitted for consideration and approved by the Research Ethics Committee of the Pontifical Catholic University of Rio de Janeiro (PUC-Rio) 045/2020 – Protocol 83/2020. It is expected that the device developed will contribute to the technological advancement of the tools used in the health sector.

[pt] Projéteis de chumbo (não-ferromagnéticos) são corpos estranhos comuns na prática médica. Métodos convencionais de localização utilizam radiações ionizantes, impondo riscos à saúde e procedimentos que duram várias horas e tipicamente terminam malsucedidos. Mapas de campos magnéticos obtidos não-invasivamente e inocuamente com SQUIDs beneficiam a localização de agulhas metálicas ferromagnéticas, reduzindo o tempo de remoção bem-sucedida de 6 horas para 10 minutos. SQUIDs são os magnetômetros mais sensíveis, entretanto requerem temperaturas criogênicas, levando a altos custos e baixa portabilidade que impedem a difusão do uso clínico. O objetivo é desenvolver um dispositivo para localizar corpos estranhos metálicos não-ferromagnéticos visando remoção cirúrgica, respeitando requerimentos de projeto: alta sensibilidade, inocuidade, não-invasividade, baixo custo, segurança, portabilidade, facilidade de uso e operação em temperatura ambiente. Sensores GMR e GMI são considerados alternativas mais adequadas. Modelos teóricos de eletrodinâmica clássica aplicados às correntes parasitas servem como base. Dois sistemas eletrônicos são desenvolvidos em configuração gradiométrica para remover interferência ambiente, usando elementos sensores GMR e GMI disponíveis comercialmente. O desempenho é obtido com resultados de simulações, provando a capacidade de detecção de níveis esperados de densidade de fluxo magnético para certos raios de projéteis e distâncias. O Sistema GMI é mais qualificado, sua mais alta sensibilidade e melhor resolução favorecem maiores faixas de medição, inocuidade, segurança e facilidade de uso. Os resultados demonstram a viabilidade dos elementos sensores GMI nessa aplicação. Os benefícios de baixo custo, maior portabilidade e segurança facilitam a utilização clínica de técnicas de localização para corpos estranhos metálicos não-ferromagnéticos mais inócuas e efetivas. / [en] Lead projectiles (non-ferromagnetic) are common foreign bodies in the medical practice. Conventional means of location use ionizing radiation, pose health risks and lead to procedures that last several hours, typically ending unsuccessfully. Magnetic field maps obtained non-invasively and innocuously with SQUIDs benefit the location of ferromagnetic metallic needles, reducing the time of successful removal from 6 hours to 10 minutes. SQUIDs are currently the most sensitive magnetometers, however require cryogenic temperatures, leading to high cost and low portability which prevent widespread clinical use. The objective is to design a device for locating non-ferromagnetic metallic foreign bodies for surgical removal, respecting project requirements of: high sensitivity, innocuousness, non-invasiveness, low cost, safety, portability, ease of use and room temperature operation. GMR and GMI sensors are considered as more suitable alternatives. Classical electrodynamics theoretical models applied to eddy currents induction serve as framework. Two electronic location systems are developed in gradiometric configuration to remove environmental interference, using commercially available GMR and GMI sensor elements. System performance is obtained from simulation results, demonstrating the capability of detecting the magnetic flux density levels expected under certain projectile radii and distances. The GMI system is more qualified, as its higher sensitivity and improved resolution favors larger measurement ranges, innocuousness, safety and ease of use. The results prove the viability of using GMI sensor elements in this application. The benefits of lower cost, higher portability and safety facilitate the clinical use of more innocuous and effective location techniques for non-ferromagnetic metallic foreign bodies.

No description available.

Neointimal hyperplasia (NIH) is a cardiovascular disease characterised by increased smooth muscle cell (SMC) inflammation and proliferation. Suppressor of cytokine signalling 3 (SOCS3) limits Janus kinase (JAK)/signal transducer and activator of transcription (STAT) pathways involved in vascular remodelling but is limited by its short biological half-life. Therefore, mutation of all 9 Lys residues that are potential sites of ubiquitylation to Arg should produce a mutated SOCS3 resistant to ubiquitin-mediated proteasomal degradation (“Lys-less” SOCS3). This study hypothesise that enhancing SOCS3 stability and limiting JAK/STAT signalling may provide sustained inhibition of the vascular remodelling in NIH. Lentiviral transduction of WT and Lys-less SOCS3 in human saphenous vein (HSVSMCs) was highly efficient after 48 hours (>97%) and was sustained over 2 weeks. Lys-less SOCS3 was resistant to ubiquitylation contrary to WT-transduced HSVECs, and Lys-less SOCS3 was more stable (t1/2=4h) than WT (t1/2<4h) (n=6, P<0.001) in HSVSMCs. In HSVSMCs, both Lys-less SOCS3 and WT inhibited sIL-6Rα/IL-6 mediated STAT3 activation but not extracellular signal regulated protein kinase 1/2 (ERK1/2) by 80±7% (Lys-lessSOCS3/pSTAT3) and 74±6% (WT/pSTAT3) (n=3, P<0.05) and similarly inhibited PDGF-mediated STAT3 activation but not ERK1/2 by 67±17% (Lys-less SOCS3/pSTAT3) and 72±18% (WT/pSTAT3) (n=3, P<0.05). Functionally, Lys-less SOCS3 and WT were equivalent in inhibiting sIL-6Rα/IL-6 and PDGF-induced proliferation, whilst having no effects on PDGF-induced migration in HSVSMCs. Lys-less SOCS3 can be successfully transduced into primary HSVSMCs. It is more stable than WT yet retains its functional ability to ameliorate pro-inflammatory signalling and SMC proliferation, making it an attractive option for developing treatment of NIH. / University of Botswana

[EN] This work is focused on problems (like automatic speech recognition (ASR) and handwritten text recognition (HTR)) that: 1) can be represented (at least approximately) in terms of one-dimensional sequences, and 2) solving these problems entails breaking the observed sequence down into segments which are associated to units taken from a finite repertoire. The required segmentation and classification tasks are so intrinsically interrelated ("Sayre's Paradox") that they have to be performed jointly. We have been inspired by what some works call the "successful trilogy", which refers to the synergistic improvements obtained when considering: - a good formalization framework and powerful algorithms; - a clever design and implementation taking the best profit of hardware; - an adequate preprocessing and a careful tuning of all heuristics. We describe and study "two stage generative models" (TSGMs) comprising two stacked probabilistic generative stages without reordering. This model not only includes Hidden Markov Models (HMMs, but also "segmental models" (SMs). "Two stage decoders" may be deduced by simply running a TSGM in reversed way, introducing non determinism when required: 1) A directed acyclic graph (DAG) is generated and 2) it is used together with a language model (LM). One-pass decoders constitute a particular case. A formalization of parsing and decoding in terms of semiring values and language equations proposes the use of recurrent transition networks (RTNs) as a normal form for Context Free Grammars (CFGs), using them in a parsing-as-composition paradigm, so that parsing CFGs result in a slight extension of regular ones. Novel transducer composition algorithms have been proposed that can work with RTNs and can deal with null transitions without resorting to filter-composition even in the presence of null transitions and non-idempotent semirings. A review of LMs is described and some contributions mainly focused on LM interfaces, LM representation and on the evaluation of Neural Network LMs (NNLMs) are provided. A review of SMs includes the combination of generative and discriminative segmental models and general scheme of frame emission and another one of SMs. Some fast cache-friendly specialized Viterbi lexicon decoders taking profit of particular HMM topologies are proposed. They are able to manage sets of active states without requiring dictionary look-ups (e.g. hashing). A dataflow architecture allowing the design of flexible and diverse recognition systems from a little repertoire of components has been proposed, including a novel DAG serialization protocol. DAG generators can take over-segmentation constraints into account, make use SMs other than HMMs, take profit of the specialized decoders proposed in this work and use a transducer model to control its behavior making it possible, for instance, to use context dependent units. Relating DAG decoders, they take profit of a general LM interface that can be extended to deal with RTNs. Some improvements for one pass decoders are proposed by combining the specialized lexicon decoders and the "bunch" extension of the LM interface, including an adequate parallelization. The experimental part is mainly focused on HTR tasks on different input modalities (offline, bimodal). We have proposed some novel preprocessing techniques for offline HTR which replace classical geometrical heuristics and make use of automatic learning techniques (neural networks). Experiments conducted on the IAM database using this new preprocessing and HMM hybridized with Multilayer Perceptrons (MLPs) have obtained some of the best results reported for this reference database. Among other HTR experiments described in this work, we have used over-segmentation information, tried lexicon free approaches, performed bimodal experiments and experimented with the combination of hybrid HMMs with holistic classifiers. / [ES] Este trabajo se centra en problemas (como reconocimiento automático del habla (ASR) o de escritura manuscrita (HTR)) que cumplen: 1) pueden representarse (quizás aproximadamente) en términos de secuencias unidimensionales, 2) su resolución implica descomponer la secuencia en segmentos que se pueden clasificar en un conjunto finito de unidades. Las tareas de segmentación y de clasificación necesarias están tan intrínsecamente interrelacionadas ("paradoja de Sayre") que deben realizarse conjuntamente. Nos hemos inspirado en lo que algunos autores denominan "La trilogía exitosa", refereido a la sinergia obtenida cuando se tiene: - un buen formalismo, que dé lugar a buenos algoritmos; - un diseño e implementación ingeniosos y eficientes, que saquen provecho de las características del hardware; - no descuidar el "saber hacer" de la tarea, un buen preproceso y el ajuste adecuado de los diversos parámetros. Describimos y estudiamos "modelos generativos en dos etapas" sin reordenamientos (TSGMs), que incluyen no sólo los modelos ocultos de Markov (HMM), sino también modelos segmentales (SMs). Se puede obtener un decodificador de "dos pasos" considerando a la inversa un TSGM introduciendo no determinismo: 1) se genera un grafo acíclico dirigido (DAG) y 2) se utiliza conjuntamente con un modelo de lenguaje (LM). El decodificador de "un paso" es un caso particular. Se formaliza el proceso de decodificación con ecuaciones de lenguajes y semianillos, se propone el uso de redes de transición recurrente (RTNs) como forma normal de gramáticas de contexto libre (CFGs) y se utiliza el paradigma de análisis por composición de manera que el análisis de CFGs resulta una extensión del análisis de FSA. Se proponen algoritmos de composición de transductores que permite el uso de RTNs y que no necesita recurrir a composición de filtros incluso en presencia de transiciones nulas y semianillos no idempotentes. Se propone una extensa revisión de LMs y algunas contribuciones relacionadas con su interfaz, con su representación y con la evaluación de LMs basados en redes neuronales (NNLMs). Se ha realizado una revisión de SMs que incluye SMs basados en combinación de modelos generativos y discriminativos, así como un esquema general de tipos de emisión de tramas y de SMs. Se proponen versiones especializadas del algoritmo de Viterbi para modelos de léxico y que manipulan estados activos sin recurrir a estructuras de tipo diccionario, sacando provecho de la caché. Se ha propuesto una arquitectura "dataflow" para obtener reconocedores a partir de un pequeño conjunto de piezas básicas con un protocolo de serialización de DAGs. Describimos generadores de DAGs que pueden tener en cuenta restricciones sobre la segmentación, utilizar modelos segmentales no limitados a HMMs, hacer uso de los decodificadores especializados propuestos en este trabajo y utilizar un transductor de control que permite el uso de unidades dependientes del contexto. Los decodificadores de DAGs hacen uso de un interfaz bastante general de LMs que ha sido extendido para permitir el uso de RTNs. Se proponen también mejoras para reconocedores "un paso" basados en algoritmos especializados para léxicos y en la interfaz de LMs en modo "bunch", así como su paralelización. La parte experimental está centrada en HTR en diversas modalidades de adquisición (offline, bimodal). Hemos propuesto técnicas novedosas para el preproceso de escritura que evita el uso de heurísticos geométricos. En su lugar, utiliza redes neuronales. Se ha probado con HMMs hibridados con redes neuronales consiguiendo, para la base de datos IAM, algunos de los mejores resultados publicados. También podemos mencionar el uso de información de sobre-segmentación, aproximaciones sin restricción de un léxico, experimentos con datos bimodales o la combinación de HMMs híbridos con reconocedores de tipo holístico. / [CA] Aquest treball es centra en problemes (com el reconeiximent automàtic de la parla (ASR) o de l'escriptura manuscrita (HTR)) on: 1) les dades es poden representar (almenys aproximadament) mitjançant seqüències unidimensionals, 2) cal descompondre la seqüència en segments que poden pertanyer a un nombre finit de tipus. Sovint, ambdues tasques es relacionen de manera tan estreta que resulta impossible separar-les ("paradoxa de Sayre") i s'han de realitzar de manera conjunta. Ens hem inspirat pel que alguns autors anomenen "trilogia exitosa", referit a la sinèrgia obtinguda quan prenim en compte: - un bon formalisme, que done lloc a bons algorismes; - un diseny i una implementació eficients, amb ingeni, que facen bon us de les particularitats del maquinari; - no perdre de vista el "saber fer", emprar un preprocés adequat i fer bon us dels diversos paràmetres. Descrivim i estudiem "models generatiu amb dues etapes" sense reordenaments (TSGMs), que inclouen no sols inclouen els models ocults de Markov (HMM), sinò també models segmentals (SM). Es pot obtindre un decodificador "en dues etapes" considerant a l'inrevés un TSGM introduint no determinisme: 1) es genera un graf acíclic dirigit (DAG) que 2) és emprat conjuntament amb un model de llenguatge (LM). El decodificador "d'un pas" en és un cas particular. Descrivim i formalitzem del procés de decodificació basada en equacions de llenguatges i en semianells. Proposem emprar xarxes de transició recurrent (RTNs) com forma normal de gramàtiques incontextuals (CFGs) i s'empra el paradigma d'anàlisi sintàctic mitjançant composició de manera que l'anàlisi de CFGs resulta una lleugera extensió de l'anàlisi de FSA. Es proposen algorismes de composició de transductors que poden emprar RTNs i que no necessiten recorrer a la composició amb filtres fins i tot amb transicions nul.les i semianells no idempotents. Es proposa una extensa revisió de LMs i algunes contribucions relacionades amb la seva interfície, amb la seva representació i amb l'avaluació de LMs basats en xarxes neuronals (NNLMs). S'ha realitzat una revisió de SMs que inclou SMs basats en la combinació de models generatius i discriminatius, així com un esquema general de tipus d'emissió de trames i altre de SMs. Es proposen versions especialitzades de l'algorisme de Viterbi per a models de lèxic que permeten emprar estats actius sense haver de recórrer a estructures de dades de tipus diccionari, i que trauen profit de la caché. S'ha proposat una arquitectura de flux de dades o "dataflow" per obtindre diversos reconeixedors a partir d'un xicotet conjunt de peces amb un protocol de serialització de DAGs. Descrivim generadors de DAGs capaços de tindre en compte restriccions sobre la segmentació, emprar models segmentals no limitats a HMMs, fer us dels decodificadors especialitzats proposats en aquest treball i emprar un transductor de control que permet emprar unitats dependents del contexte. Els decodificadors de DAGs fan us d'una interfície de LMs prou general que ha segut extesa per permetre l'ús de RTNs. Es proposen millores per a reconeixedors de tipus "un pas" basats en els algorismes especialitzats per a lèxics i en la interfície de LMs en mode "bunch", així com la seua paral.lelització. La part experimental està centrada en el reconeiximent d'escriptura en diverses modalitats d'adquisició (offline, bimodal). Proposem un preprocés d'escriptura manuscrita evitant l'us d'heurístics geomètrics, en el seu lloc emprem xarxes neuronals. S'han emprat HMMs hibridats amb xarxes neuronals aconseguint, per a la base de dades IAM, alguns dels millors resultats publicats. També podem mencionar l'ús d'informació de sobre-segmentació, aproximacions sense restricció a un lèxic, experiments amb dades bimodals o la combinació de HMMs híbrids amb classificadors holístics. / España Boquera, S. (2016). Contributions to the joint segmentation and classification of sequences (My two cents on decoding and handwriting recognition) [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/62215 / Premios Extraordinarios de tesis doctorales

Dans ce travail, une approche de reconnaissance de l’orateur en utilisant un microphone de contact est développée et présentée. L'élément passif de contact est construit à partir d'un matériau piézoélectrique. La position du transducteur piézoélectrique sur le cou de l'individu peut affecter grandement la qualité du signal recueilli et par conséquent les informations qui en sont extraites. Ainsi, le milieu multicouche dans lequel les vibrations des cordes vocales se propagent avant d'être détectées par le transducteur est modélisé. Le meilleur emplacement sur le cou de l’individu pour attacher un élément transducteur particulier est déterminé en mettant en œuvre des techniques de simulation Monte Carlo et, par conséquent, les résultats de la simulation sont vérifiés en utilisant des expériences réelles. La reconnaissance est basée sur le signal généré par les vibrations des cordes vocales lorsqu'un individu parle et non sur le signal vocal à la sortie des lèvres qui est influencé par les résonances dans le conduit vocal. Par conséquent, en raison de la nature variable du signal recueilli, l'analyse a été effectuée en appliquant la technique de transformation de Fourier à court terme pour décomposer le signal en ses composantes de fréquence. Ces fréquences représentent les vibrations des cordes vocales (50-1000 Hz). Les caractéristiques en termes d'intervalle de fréquences sont extraites du spectrogramme résultant. Ensuite, un vecteur 1-D est formé à des fins d'identification. L'identification de l’orateur est effectuée en utilisant deux critères d'évaluation qui sont la mesure de la similarité de corrélation et l'analyse en composantes principales (ACP) en conjonction avec la distance euclidienne. Les résultats montrent qu'un pourcentage élevé de reconnaissance est atteint et que la performance est bien meilleure que de nombreuses techniques existantes dans la littérature. / In this work, a speaker recognition approach using a contact microphone is developed and presented. The contact passive element is constructed from a piezoelectric material. In this context, the position of the piezoelectric transducer on the individual’s neck may greatly affect the quality of the collected signal and consequently the information extracted from it. Thus, the multilayered medium in which the sound propagates before being detected by the transducer is modeled. The best location on the individual’ neck to place a particular transducer element is determined by implementing Monte Carlo simulation techniques and consequently, the simulation results are verified using real experiments. The recognition is based on the signal generated from the vocal cords’ vibrations when an individual is speaking and not on the vocal signal at the output of the lips that is influenced by the resonances in the vocal tract. Therefore, due to the varying nature of the collected signal, the analysis was performed by applying the Short Term Fourier Transform technique to decompose the signal into its frequency components. These frequencies represent the vocal folds’ vibrations (50-1000 Hz). The features in terms of frequencies’ interval are extracted from the resulting spectrogram. Then, a 1-D vector is formed for identification purposes. The identification of the speaker is performed using two evaluation criteria, namely, the correlation similarity measure and the Principal Component Analysis (PCA) in conjunction with the Euclidean distance. The results show that a high percentage of recognition is achieved and the performance is much better than many existing techniques in the literature.