Le nombre croissant d'appareils de diffusion de contenus audio pose le problème de partager le même espace physique sans partager lemême espace sonore. Les Sound Zones rendent possible la reproduction de programmes audio indépendants et spatialement séparés, àpartir d'un ensemble de haut-parleurs et de méthodes de reproduction de champs sonores. Le problème est alors décomposé en deuxzones : la Bright zone, où le contenu doit être reproduit et la Dark zone, où il doit être annulé. De nombreuses méthodes existent pourrésoudre ce problème, mais la plupart ne traite que le cas d'auditeurs en position statique. Elles s'appuient sur la résolution directe desméthodes d'optimisation adaptative, telle que la méthode de Pressure Matching (PM). Or, pour des utilisateurs en mouvement, cesméthodes ont un coût de calcul trop élevé, rendant impossible leur application à un problème dynamique. Le but de cette thèse est dedévelopper une solution présentant une complexité compatible avec un contrôle dynamique des Sound Zones, tout en conservant lesperformances des méthodes conventionnelles. Sous l'hypothèse que les déplacements sont lents, une résolution itérative du problème PMest proposée et évaluée. Les algorithmes LMS, NLMS et APA sont comparés sur la base de simulations en champ libre. La méthode LMSs'avère la plus avantageuse en termes de complexité, mais elle souffre d'une erreur de reproduction. Un effet mémoire limitant la réactivitédes algorithmes est aussi mis en évidence. Il est corrigé en implémentant une variante introduisant un facteur d'oubli (Variable LeakyLMS ou VLLMS). / The growing number of audio devices raises the problem of sharing the same physical space without sharing the same sound space. SoundZones make it possible to play independent and spatially separated audio programs by loudspeaker array in combination with sound fieldreproduction methods. The problem is then split into two zones: the Bright zone, where the audio content must be reproduced and theDark zone, where it must be cancelled. There are many methods available to solve this problem, but most only deal with auditors in astatic position. They are based on the direct resolution of adaptive optimization methods, such as the Pressure Matching (PM) method.However, for moving users, these methods have a too high computation cost, making it impossible to apply them to a dynamic problem.The aim of this thesis is to develop a solution offering a level of complexity compatible with a dynamic control of Sound Zones, whilemaintening the performance of conventional methods. Under the assumption that displacements are slow, an iterative resolution of the PMproblem is proposed and assessed. The LMS, NLMS and APA algorithms are compared on the basis of free field simulations. The LMSmethod is the most advantageous in terms of complexity, but it suffers from a reproduction error. A memory effect limiting the reactivityof the algorithms is also highlighted. It is corrected by implementing a leaky variant (Variable Leaky LMS or VLLMS) introducing aforgetting factor.

La problématique traitée dans cette thèse vise à développer une stratégie efficace de détection et de classification des impacts en présence d'incertitudes de modélisation du robot et de son environnement et en utilisant un nombre minimal de capteurs, notamment en l'absence de capteur d’effort.La première partie de la thèse porte sur la détection d'un impact pouvant avoir lieu à n'importe quel endroit du bras robotique et à n'importe quel moment de sa trajectoire. Les méthodes de détection d’impacts sont généralement basées sur un modèle dynamique du système, ce qui les rend sujettes au compromis entre sensibilité de détection et robustesse aux incertitudes de modélisation. A cet égard, une méthodologie quantitative a d'abord été mise au point pour rendre explicite la contribution des erreurs induites par les incertitudes de modèle. Cette méthodologie a été appliquée à différentes stratégies de détection, basées soit sur une estimation directe du couple extérieur, soit sur l'utilisation d'observateurs de perturbation, dans le cas d’une modélisation parfaitement rigide ou à articulations flexibles. Une comparaison du type et de la structure des erreurs qui en découlent et de leurs conséquences sur la détection d'impacts en a été déduite. Dans une deuxième étape, de nouvelles stratégies de détection d'impacts ont été conçues: les effets dynamiques des impacts sont isolés en déterminant la marge d'erreur maximale due aux incertitudes de modèle à l’aide d’une approche stochastique.Une fois l'impact détecté et afin de déclencher la réaction post-impact du robot la plus appropriée, la deuxième partie de la thèse aborde l'étape de classification. En particulier, la distinction entre un contact intentionnel (l'opérateur interagit intentionnellement avec le robot, par exemple pour reconfigurer la tâche) et un contact non-désiré (un sujet humain heurte accidentellement le robot), ainsi que la localisation du contact sur le robot, est étudiée en utilisant des techniques d'apprentissage supervisé et plus spécifiquement des réseaux de neurones feedforward. La généralisation à plusieurs sujet humains et à différentes trajectoires du robot a été étudiée. / The present thesis aims to develop an efficient strategy for impact detection and classification in the presence of modeling uncertainties of the robot and its environment and using a minimum number of sensors, in particular in the absence of force/torque sensor.The first part of the thesis deals with the detection of an impact that can occur at any location along the robot arm and at any moment during the robot trajectory. Impact detection methods are commonly based on a dynamic model of the system, making them subject to the trade-off between sensitivity of detection and robustness to modeling uncertainties. In this respect, a quantitative methodology has first been developed to make explicit the contribution of the errors induced by model uncertainties. This methodology has been applied to various detection strategies, based either on a direct estimate of the external torque or using disturbance observers, in the perfectly rigid case or in the elastic-joint case. A comparison of the type and structure of the errors involved and their consequences on the impact detection has been deduced. In a second step, novel impact detection strategies have been designed: the dynamic effects of the impacts are isolated by determining the maximal error range due to modeling uncertainties using a stochastic approach.Once the impact has been detected and in order to trigger the most appropriate post-impact robot reaction, the second part of the thesis focuses on the classification step. In particular, the distinction between an intentional contact (the human operator intentionally interacts with the robot, for example to reconfigure the task) and an undesired contact (a human subject accidentally runs into the robot), as well as the localization of the contact on the robot, is investigated using supervised learning techniques and more specifically feedforward neural networks. The challenge of generalizing to several human subjects and robot trajectories has been investigated.

It is known that a matched filter is optimal for a signal corrupted by Gaussian noise. In a wireless environment, the received signal may be corrupted by Gaussian noise and a variety of other channel disturbances: cochannel interference, multiple access interference, large and small-scale fading, etc. Adaptive filtering is the usual approach to mitigating this channel distortion. Existing adaptive filtering techniques usually attempt to minimize the mean square error (MSE) of some aspect of the received signal, with respect to the desired aspect of that signal. Adaptive minimization of MSE does not always guarantee minimization of bit error rate (BER). The main focus of this research involves estimation of the probability density function (PDF) of the received signal; this PDF estimate is used to adaptively determine a solution that minimizes BER. To this end, a new adaptive procedure called the Minimum BER Estimation (MBE) algorithm has been developed. MBE shows improvement over the Least Mean Squares (LMS) algorithm for most simulations involving interference and in some multipath situations. Furthermore, the new algorithm is more robust than LMS to changes in algorithm parameters such as stepsize and window width. / Master of Science

No description available.

No description available.

[pt] Esta tese é composta de três artigos em que a ligação entre eles são modelos estatísticos com parametros variantes no tempo. Todos os artigos adotam um arcabouço que utiliza um mecanismo guiado pelos dados para a atualização dos parâmetros dos modelos. O primeiro explora a aplicação de uma nova classe de modelos de séries temporais não Gaussianas denominada modelos Generalized Autegressive Scores (GAS). Nessa classe de modelos, os parâmetros são atualizados utilizando o score da densidade preditiva. Motivamos o uso de modelos GAS simulando cenários conjuntos de fator de capacidade eólico. Nos últimos dois artigos, o gradiente descentente estocástico (SGD) é adotado para atualizar os parâmetros que variam no tempo. Tal metodologia utiliza a derivada de uma função custo especificada pelo usuário para guiar a otimização. A estrutura desenvolvida foi projetada para ser aplicada em um contexto de fluxo de dados contínuo, portanto, técnicas de filtragem adaptativa são exploradas para levar em consideração o concept-drift. Exploramos esse arcabouço com aplicações em segurança cibernética e infra-estrutura instrumentada. / [en] This thesis is composed of three papers in which the common ground among them is statistical models with time-varying parameters. All of them adopt a framework that uses a data-driven mechanism to update its coefficients. The first paper explores the application of a new class of non-Gaussian time series framework named Generalized Autoregressive Scores (GAS) models. In this class of models the parameters are updated using the score of the predictive density. We motivate the use of GAS models by simulating joint scenarios of wind power generation. In the last two papers, Stochastic Gradient Descent (SGD) is adopted to update time-varying parameters. This methodology uses the derivative of a user specified cost function to drive the optimization. The developed framework is designed to be applied in a streaming data context, therefore adaptive filtering techniques are explored to account for concept-drift.We explore this framework on cyber-security and instrumented infrastructure applications.

The great social and economic impact of earthquakes has made necessary the development of novel structural health monitoring (SHM) solutions for increasing the level of structural safety and assessment. SHM is the process of comparing the current state of a structure’s condition relative to a healthy baseline state to detect the existence, location, and degree of likely damage during or after a damaging input, such as an earthquake. Many SHM algorithms have been proposed in the literature. However, a large majority of these algorithms cannot be implemented in real time. Therefore, their results would not be available during or immediately after a major event for urgent post-event response and decision making. Further, these off-line techniques are not capable of providing the input information required for structural control systems for damage mitigation. The small number of real-time SHM (RT-SHM) methods proposed in the past, resolve these issues. However, these approaches have significant computational complexity and typically do not manage nonlinear cases directly associated with relevant damage metrics. Finally, many available SHM methods require full structural response measurement, including velocities and displacements, which are typically difficult to measure. All these issues make implementation of many existing SHM algorithms very difficult if not impossible. This thesis proposes simpler, more suitable algorithms utilising a nonlinear Bouc-Wen hysteretic baseline model for RT-SHM of a large class of nonlinear hysteretic structures. The RT-SHM algorithms are devised so that they can accommodate different levels of the availability of design data or measured structural responses, and therefore, are applicable to both existing and new structures. The second focus of the thesis is on developing a high-speed, high-resolution, seismic structural displacement measurement sensor to enable these methods and many other SHM approaches by using line-scan cameras as a low-cost and powerful means of measuring structural displacements at high sampling rates and high resolution. Overall, the results presented are thus significant steps towards developing smart, damage-free structures and providing more reliable information for post-event decision making.

La séparation de signaux corrélés à partir de leurs combinaisons linéaires est une tâche difficile et possède plusieurs applications en traitement du signal. Nous étudions deux problèmes, à savoir la séparation aveugle de sources parcimonieuses et le filtrage adaptatif des réflexions multiples en acquisition sismique. Un intérêt particulier est porté sur les mélanges convolutifs : pour ces deux problèmes, des filtres à réponses impulsionnelles finies peuvent être estimés afin de récupérer les signaux désirés.Pour les modèles de mélange instantanés et convolutifs, nous donnons les conditions nécessaires et suffisantes pour l'extraction et la séparation exactes de sources parcimonieuses en utilisant la pseudo-norme L0 comme une fonction de contraste. Des équivalences entre l'analyse en composantes parcimonieuses et l'analyse en composantes disjointes sont examinées.Pour la soustraction adaptative des réflexions sismiques, nous discutons les limites des méthodes basées sur l'analyse en composantes indépendantes et nous soulignons l'équivalence avec les méthodes basées sur les normes Lp. Nous examinons de quelle manière les paramètres de régularisation peuvent être plus décisifs pour l'estimation des primaires. Enfin, nous proposons une amélioration de la robustesse de la soustraction adaptative en estimant les filtres adaptatifs directement dans le domaine des curvelets. Les coûts en calcul et en mémoire peuvent être atténués par l'utilisation de la transformée en curvelet discrète et uniforme. / The recovery of correlated signals from their linear combinations is a challenging task and has many applications in signal processing. We focus on two problems that are the blind separation of sparse sources and the adaptive subtraction of multiple events in seismic processing. A special focus is put on convolutive mixtures: for both problems, finite impulse response filters can indeed be estimated for the recovery of the desired signals.For instantaneous and convolutive mixing models, we address the necessary and sufficient conditions for the exact extraction and separation of sparse sources by using the L0 pseudo-norm as a contrast function. Equivalences between sparse component analysis and disjoint component analysis are investigated.For adaptive multiple subtraction, we discuss the limits of methods based on independent component analysis and we highlight equivalence with Lp-norm-based methods. We investigate how other regularization parameters may have more influence on the estimation of the desired primaries. Finally, we propose to improve the robustness of adaptive subtraction by estimating the extracting convolutive filters directly in the curvelet domain. Computation and memory costs are limited by using the uniform discrete curvelet transform.

Um Sistema Elétrico de Potência (SEP) está susceptível à presença de diversas fontes de distúrbios que prejudicam a Qualidade da Energia Elétrica (QEE). Desta forma, as suas tensões e/ou correntes podem conter m´múltiplos distúrbios com ocorrência simultânea. Este trabalho apresenta uma metodologia para decomposição do sinal medido em componentes que estimem as formas de onda dos distúrbios individuais quando da ocorrência de m´múltiplos distúrbios, com o posterior reconhecimento de cada um deles. A Análise de Componentes Independentes (ICA) é utilizada como principal ferramenta na etapa de decomposição dos distúrbios. A ICA é originalmente uma t´técnica aplicada em análise multivariada de dados, o que significa que ela necessita de medições realizadas por múltiplos sensores dispostos em diferentes posições de um sistema. No entanto, este trabalho propõe a sua aplicação tendo disponível apenas um sinal medido. Para tanto, são propostos dois métodos para produzir a diversidade necessária para a t´técnica funcionar adequadamente. É demonstrado que ambos os métodos equivalem a um banco de filtros lineares adaptativos capaz de realizar a separação não-supervisionada de múltiplos distúrbios independentes e que sejam espectralmente disjuntos. Por fim, é proposto um sistema de classificação que utiliza Redes Neurais Artificiais (RNAs) para identificar os distúrbios decompostos pela etapa anterior. A metodologia completa é avaliada por meio de testes utilizando dados sintéticos e reais, alcançando resultados altamente satisfatórios para decomposição de sinais contendo múltiplos distúrbios e taxas de acerto globais dos classificadores superiores a 97% / The power system is susceptible to the presence of several sources of disturbances that harm the power quality. In this sense, its voltages and/or currents may contain multiple disturbances with simultaneous occurrence. This work presents a methodology that decomposes the measured signal in components which estimate the waveforms of the individual disturbances followed by their recognition when a multiple disturbance situation occurs. The Independent Component Analysis (ICA) is the main tool in the disturbance decomposition stage. The ICA is originally a technique applied in multivariate data analysis, which means that it requires measurements from multiple sensors allocated in different positions of the system. However, this work proposes its application for a single measured signal available. For this, two methods were developed in order to provide the required diversity to the ICA technique. It is demonstrated that both methods are equivalent to an adaptive linear filter bank capable to perform an unsupervised separation of multiple independent disturbances, if they are spectrally disjoint. A classification system based on artificial neural networks is proposed to identify the disturbances decomposed by the previous stage. The complete system is tested using synthetic and actual data, presenting highly satisfactory results for the decomposition of signals containing multiple disturbances, and precision for the classification task above 97%

Neste trabalho, são propostos e analisados algoritmos autodidatas eficientes para a equalização de canais de comunicação, considerando a transmissão de sinais QAM (quadrature amplitude modulation). Suas funções de erro são construídas de forma a fazer com que o erro de estimação seja igual a zero nas coordenadas dos símbolos da constelação. Essa característica os possibilita ter um desempenho similar ao de um algoritmo de equalização supervisionada como o NLMS (normalized least mean-square), independentemente da ordem da constelação QAM. Verifica-se analiticamente que, sob certas condições favoráveis para a equalização, os vetores de coeficientes dos algoritmos propostos e a correspondente solução de Wiener são colineares. Além disso, usando a informação da estimativa do símbolo transmitido e de seus símbolos vizinhos, esquemas de baixo custo computacional são propostos para aumentar a velocidade de convergência dos algoritmos. No caso do algoritmo baseado no critério do módulo constante, evita-se sua divergência através de um mecanismo que descarta estimativas inconsistentes dos símbolos transmitidos. Adicionalmente, apresenta-se uma análise de rastreio (tracking), que permite obter expressões analíticas para o erro quadrático médio em excesso dos algoritmos propostos em ambientes estacionários e não-estacionários. Através dessas expressões, verifica-se que com sobreamostragem, ausência de ruído e ambiente estacionário, os algoritmos propostos podem alcançar a equalização perfeita, independentemente da ordem da constelação QAM. Os algoritmos são estendidos para a adaptação conjunta dos filtros direto e de realimentação do equalizador de decisão realimentada, levando-se em conta um mecanismo que evita soluções degeneradas. Resultados de simulação sugerem que a utilização dos esquemas aqui propostos pode ser vantajosa na recuperação de sinais QAM, fazendo com que seja desnecessário o chaveamento para o algoritmo de decisão direta. / In this work, we propose efficient blind algorithms for equalization of communication channels, considering the transmission of QAM (quadrature amplitude modulation) signals. Their error functions are constructed in order to make the estimation error equal to zero at the coordinates of the constellation symbols. This characteristic enables the proposed algorithms to have a similar performance to that of a supervised equalization algorithm as the NLMS (normalized least mean-square), independently of the QAM order. Under some favorable conditions, we verify analytically that the coefficient vector of the proposed algorithms are collinear with the Wiener solution. Furthermore, using the information of the symbol estimate in conjunction with its neighborhood, we propose schemes of low computational cost in order to improve their convergence rate. The divergence of the constant-modulus based algorithm is avoided by using a mechanism, which disregards nonconsistent estimates of the transmitted symbols. Additionally, we present a tracking analysis in which we obtain analytical expressions for the excess mean-square error in stationary and nonstationary environments. From these expressions, we verify that using a fractionally-spaced equalizer in a noiseless stationary environment, the proposed algorithms can achieve perfect equalization, independently of the QAM order. The algorithms are extended to jointly adapt the feedforward and feedback filters of the decision feedback equalizer, taking into account a mechanism to avoid degenerative solutions. Simulation results suggest that the proposed schemes may be advantageously used to recover QAM signals and make the switching to the decision direct mode unnecessary.