This thesis presents a research work based on an integrated multi-modal approach for sports video indexing and retrieval. By combining specific features extractable from multiple (audio-visual) modalities, generic structure and specific events can be detected and classified. During browsing and retrieval, users will benefit from the integration of high-level semantic and some descriptive mid-level features such as whistle and close-up view of player(s). The main objective is to contribute to the three major components of sports video indexing systems. The first component is a set of powerful techniques to extract audio-visual features and semantic contents automatically. The main purposes are to reduce manual annotations and to summarize the lengthy contents into a compact, meaningful and more enjoyable presentation. The second component is an expressive and flexible indexing technique that supports gradual index construction. Indexing scheme is essential to determine the methods by which users can access a video database. The third and last component is a query language that can generate dynamic video summaries for smart browsing and support user-oriented retrievals.

The development of self-powered wireless sensor networks for structural and machinery health monitoring has attracted considerable attention in the research field during the last decade. Since the low-duty-cycle wireless sensor networks have significantly reduced the power requirements to the range of tens to hundreds of microwatts, it is possible to harvest environmental energy as the power supply instead of using batteries. Vibration energy harvesting using piezoelectric materials has become the most popular technique, which has a good potential to generate adequate power. However, there is a limitation for the conventional beam-shaped harvester designs in real applications due to their limited bandwidth. In order to overcome this limitation, the essential objective of this thesis is to develop harvesters with multi-resonance structures. The multi-resonance harvester with good broadband performance can achieve close resonance frequencies and relatively large power output in each vibration mode. The main tasks and contributions of this thesis are summarised as follows: • A parametric analysis is presented to determine how the modal structural and electromechanical performances of cantilevered beam harvesters are affected by two modal factors designated as mass ratio and electromechanical coupling coefficient (EMCC). The modal performance of using rectangular, convergent and divergent tapered configurations with and without extra masses are systematically analysed by geometric variation using the finite element analysis (FEA) software ABAQUS. • A modal approach using the two modal factors to evaluate the modal performance of harvesters is introduced and a configurational optimization strategy based on the modal approach is developed to pre-select the configurations of multi-resonance harvesters with better modal structural performance and close resonance frequencies in multiple modes. Using this optimization strategy obviates the need to run the full analysis at the first stage. • A novel two-layer stacked harvester, which consists of a base cantilevered beam that is connected to an upper beam by a rigid mass, is developed. By altering the dimensions and the locations of the masses, the two-layer harvester can generate two close resonance frequencies with relatively large power output. The effects of using rectangular, convergent and divergent tapered beam configurations are systematically analysed. • Multi-layer stacked harvesters with up to five layers are developed. The three-layer harvesters with different mass positions, which can generate three close resonance frequencies, are optimized using the configurational optimization strategy. • A novel doubly-clamped multi-layer harvester, which is able to generate five close resonance frequencies with relatively large power output, is developed and thoroughly analysed. • An experimental study of the multi-layer stacked harvester is presented to validate the simulated results and the configurational optimization strategy. • An experimental study of the two-layer stacked harvester using high performance single crystal piezoelectric material PIMNT is presented. The harvester using PIMNT can generate nearly 10 times larger power output and 3.5 times wider bandwidth than using PZT. Besides, by modifying the location of the piezoelectric layer, anti-resonances between two adjacent modes can be eliminated.

Dans le contexte des Sciences de l’Information et de la Technologie, un des challenges est de créer des systèmes de plus en plus petits embarquant de plus en plus d’intelligence au niveau matériel et logiciel avec des architectures communicantes de plus en plus complexes. Ceci nécessite des méthodologies robustes de conception afin de réduire le cycle de développement et la phase de prototypage. Ainsi, la conception et l’optimisation de la couche physique de communication est primordiale. La complexité de ces systèmes rend difficile leur optimisation notamment à cause de l’explosion du nombre des paramètres inconnus. Les méthodes et outils développés ces dernières années seront à terme inadéquats pour traiter les problèmes qui nous attendent. Par exemple, la propagation des ondes dans une cabine d’avion à partir des capteurs ou même d’une antenne, vers le poste de pilotage est grandement affectée par la présence de la structure métallique des sièges à l’intérieur de la cabine, voir les passagers. Il faut, donc, absolument prendre en compte cette perturbation pour prédire correctement le bilan de puissance entre l’antenne et un possible récepteur. Ces travaux de recherche portent sur les aspects théoriques et de mise en oeuvre pratique afin de proposer des outils informatiques pour le calcul rigoureux de la réflexion des champs électromagnétiques à l’intérieur de très grandes structures . Ce calcul implique la solution numérique de très grands systèmes inaccessibles par des ressources traditionnelles. La solution sera basée sur une grille de calcul et un supercalculateur. La modélisation électromagnétique des structures surdimensionnées par plusieurs méthodes numériques utilisant des nouvelles ressources informatiques, hardware et software, pour dérouler des calculs performants, représente le but de ce travail. La modélisation numérique est basée sur une approche hybride qui combine la méthode Transmission-Line Matrix (TLM) et l’approche modale. La TLM est appliquée aux volumes homogènes, tandis que l’approche modale est utilisée pour décrire les structures planaires complexes. Afin d’accélérer la simulation, une implémentation parallèle de l’algorithme TLM dans le contexte du paradigme de calcul distribué est proposé. Le sous-domaine de la structure qui est discrétisé avec la TLM est divisé en plusieurs parties appelées tâches, chacune étant calculée en parallèle par des processeurs différents. Pour accomplir le travail, les tâches communiquent entre elles au cours de la simulation par une librairie d’échange de messages. Une extension de l’approche modale avec plusieurs modes différents a été développée par l’augmentation de la complexité des structures planaires. Les résultats démontrent les avantages de la grille de calcul combinée avec l’approche hybride pour résoudre des grandes structures électriques, en faisant correspondre la taille du problème avec le nombre de ressources de calcul utilisées. L’étude met en évidence le rôle du schéma de parallélisation, cluster versus grille, par rapport à la taille du problème et à sa répartition. En outre, un modèle de prédiction a été développé pour déterminer les performances du calcul sur la grille, basé sur une approche hybride qui combine une prédiction issue d’un historique d’expériences avec une prédiction dérivée du profil de l’application. Les valeurs prédites sont en bon accord avec les valeurs mesurées. L’analyse des performances de simulation a permis d’extraire des règles pratiques pour l’estimation des ressources nécessaires pour un problème donné. En utilisant tous ces outils, la propagation du champ électromagnétique à l’intérieur d’une structure surdimensionnée complexe, telle qu’une cabine d’avion, a été effectuée sur la grille et également sur le supercalculateur. Les avantages et les inconvénients des deux environnements sont discutés. / In the context of Information Communications Technology (ICT), the major challenge is to create systems increasingly small, boarding more and more intelligence, hardware and software, including complex communicating architectures. This requires robust design methodologies to reduce the development cycle and prototyping phase. Thus, the design and optimization of physical layer communication is paramount. The complexity of these systems makes them difficult to optimize, because of the explosion in the number of unknown parameters. The methods and tools developed in past years will be eventually inadequate to address problems that lie ahead. Communicating objects will be very often integrated into cluttered environments with all kinds of metal structures and dielectric larger or smaller sizes compared to the wavelength. The designer must anticipate the presence of such barriers in the propagation channel to establish properly link budgets and an optimal design of the communicating object. For example, the wave propagation in an airplane cabin from sensors or even an antenna, towards the cockpit is greatly affected by the presence of the metal structure of the seats inside the cabin or even the passengers. So, we must absolutely take into account this perturbation to predict correctly the power balance between the antenna and a possible receiver. More generally, this topic will address the theoretical and computational electromagnetics in order to propose an implementation of informatics tools for the rigorous calculation of electromagnetic scattering inside very large structures or radiation antenna placed near oversized objects. This calculation involves the numerical solution of very large systems inaccessible by traditional resources. The solution will be based on grid computing and supercomputers. Electromagnetic modeling of oversized structures by means of different numerical methods, using new resources (hardware and software) to realize yet more performant calculations, is the aim of this work. The numerical modeling is based on a hybrid approach which combines Transmission-Line Matrix (TLM) and the mode matching methods. The former is applied to homogeneous volumes while the latter is used to describe complex planar structures. In order to accelerate the simulation, a parallel implementation of the TLM algorithm in the context of distributed computing paradigm is proposed. The subdomain of the structure which is discretized upon TLM is divided into several parts called tasks, each one being computed in parallel by different processors. To achieve this, the tasks communicate between them during the simulation by a message passing library. An extension of the modal approach to various modes has been developped by increasing the complexity of the planar structures. The results prove the benefits of the combined grid computing and hybrid approach to solve electrically large structures, by matching the size of the problem with the number of computing resources used. The study highlights the role of parallelization scheme, cluster versus grid, with respect to the size of the problem and its repartition. Moreover, a prediction model for the computing performances on grid, based on a hybrid approach that combines a historic-based prediction and an application profile-based prediction, has been developped. The predicted values are in good agreement with the measured values. The analysis of the simulation performances has allowed to extract practical rules for the estimation of the required resources for a given problem. Using all these tools, the propagation of the electromagnetic field inside a complex oversized structure such an airplane cabin, has been performed on grid and also on a supercomputer. The advantages and disadvantages of the two environments are discussed.

Structural Health Monitoring (SHM) is a non-destructive evaluation tool that assesses the functionality of structural systems that are used in the civil, mechanical and aerospace engineering practices. A much desirable objective of a SHM system is to provide a continuous monitoring service at a minimal cost with ability to identify problems even in inaccessible structural components. In this dissertation, several such approaches that utilize the measured dynamic response of structural systems are presented to detect, locate, and quantify the damages that are likely to occur in structures. In this study, the structural damage is identified as a reduction in the stiffness characteristics of the structural elements. The primary focus of this study is on the utilization of measured dynamic strains for damage identification in the framed structures which are composed of interconnected beam elements. Although linear accelerations, being more convenient to measure, are commonly used in most SHM practices, herein the strains being more sensitive to elemental damage are considered. Two different approaches are investigated and proposed to identify the structural element stiffness properties. Both approaches are mode-based, requiring first the identification of system modes from the measured strain responses followed by the identification of the element stiffness coefficients. The first approach utilizes the Eigen equation of the finite element model of the structure, while the second approach utilizes the changes caused by the damage in the structural curvature flexibilities. To reduce size of the system which is primarily determined by the number of sensors deployed for the dynamic data collection, measurement sensitivity-based sensor selection criterion is observed to be effective and thus used. The mean square values of the measurements with respect to the stiffness coefficients of the structural elements are used as the effective measures of the measurement sensitivities at different sensor locations. Numerical simulations are used to evaluate the proposed identification approaches as well as to validate the sensitivity-based optimal sensor deployment approach. / Ph. D. / All modern societies depend heavily on civil infrastructure systems such as transportation systems, power generation and transmission systems, and data communication systems for their day-to-day activities and survival. It has become extremely important that these systems are constantly watched and maintained to ensure their functionality. All these infrastructure systems utilize structural systems of different forms such as buildings, bridges, airplanes, data communication towers, etc. that carry the service and environmental loads that are imposed on them. These structural systems deteriorate over time because of natural material degradation. They can also get damaged due to excessive load demands and unknown construction deficiencies. It is necessary that condition of these structural systems is known at all times to maintain their functionality and to avoid sudden breakdowns and associated ensuing problems. This condition assessment of structural systems, now commonly known as structural health monitoring, is commonly done by visual onsite inspections manually performed at pre-decided time intervals such as on monthly and yearly basis. The length of this inspection time interval usually depends on the relative importance of the structure towards the functionality of the larger infrastructure system. This manual inspection can be highly time and resource consuming, and often ineffective in catching structural defects that are inaccessible and those that occur in between the scheduled inspection times and dates. However, the development of new sensors, new instrumentation techniques, and large data transfer and processing methods now make it possible to do this structural health monitoring on a continuous basis. The primary objective of this study is to utilize the measured dynamic or time varying strains on structural components such as beams, columns and other structural members to detect the location and level of a damage in one or more structural elements before they become serious. This detection can be done on a continuous basis by analyzing the available strain response data. This approach is expected to be especially helpful in alerting the owner of a structure by identifying the iv occurrence of a damage, if any, immediately after an unanticipated occurrence of a natural event such as a strong earthquake or a damaging wind storm.

Ce mémoire de thèse présente le développement de la méthode «statistical modal energy distribution analysis (SmEdA)» pour des systèmes vibro-acoustiques couplés. Cette méthode de calcul est basée sur le bilan énergétique dans des sous-systèmes fermés couplés, comme une structure ou une cavité. L’interaction entre de tels systèmes est décrite par des couplages entre les modes. La version initiale de SmEdA prend en compte seulement les modes qui ont une fréquence propre dans le bande d’excitation. Le travail présenté ici étudie l’effet des modes non résonants sur la réponse et identifie les cas dans lesquels un tel effet devient important. L’introduction des modes non résonants permet d’utiliser la méthode SmEdA dans des cas d’applications plus larges. En outre, une nouvelle méthode de post-traitement a été développée pour calculer des distributions d'énergie dans les sous-systèmes. Finalement, une nouvelle méthode d'approximation pour la prise en compte des modes de systèmes de grandes dimensions ou mal définis a été formulée. Toutes ces méthodes ont été comparées avec d’autres méthodes de calcul via des exemples académiques et industriels. Ainsi, la nouvelle version de SmEdA incluant le post-traitement pour obtenir des distributions d'énergie a été validé et les avantages et possibilités d'applications sont montrés. / This dissertation presents the further development of the statistical modal energy distribution analysis (SmEdA) for vibro-acoustic coupled problems. This prediction method is based on the energy balance in bounded coupled subsystems, like a structure or a cavity. The interaction between such subsystems is described by mode-to-mode coupling. The original SmEdA formulation takes into account only the modes having the eigenfrequencies within the excitation band. The present work investigates the effect of non resonant modes to the response and identifies cases in which such an effect becomes important. The inclusion of non resonant modes has thus resulted in a new SmEdA formulation which can be used in extended applications. Furthermore, a new post-processing method has been developed to predict energy distribution within subsystems. Finally a novel approximation method for handling modes of huge or ill-defined systems has been formulated. All these methods have been compared to other prediction methods via academic and industrial examples. In this way, the extended SmEdA approach including the post-processing for energy distribution has been validated and its advantages and application possibilities have been demonstrated.

L’intégration de l’électronique dans des environnements sévères d’un point de vue électromagnétique a entraîné en contrepartie l’apparition de problèmes de compatibilité électromagnétique (CEM) entre les différents systèmes. Afin d’atteindre un niveau de performance satisfaisant, des tests de sécurité et de certification sont nécessaires. Ces travaux de thèse, réalisés dans le cadre du projet SIMUCEDO (SIMUlation CEM basée sur la norme DO-160), contribuent à la modélisation du test de qualification "Bulk Current Injection" (BCI). Ce test, abordé dans la section 20 dans la norme DO-160 dédiée à l’aéronautique, est désormais obligatoire pour une très grande gamme d’équipements aéronautiques. Parmi les essais de qualification, le test BCI est l’un des plus contraignants et consommateurs du temps. Sa modélisation assure un gain de temps, et une meilleure maîtrise des paramètres qui influencent le passage des tests CEM. La modélisation du test a été décomposée en deux parties : l’équipement sous test (EST) d’une part, et la pince d’injection avec les câbles d’autre part. Dans cette thèse, seul l’EST est pris en compte. Une modélisation "boîte grise" a été proposée en associant un modèle "boîte noire" avec un modèle "extensif". Le modèle boîte noire s’appuie sur la mesure des impédances standards. Son identification se fait avec un modèle en pi. Le modèle extensif permet d’étudier plusieurs configurations de l’EST en ajustant les paramètres physiques. L’assemblage des deux modèles en un modèle boîte grise a été validé sur un convertisseur analogique-numérique (CAN). Une autre approche dénommée approche modale en fonction du mode commun (MC) et du mode différentiel (MD) a été proposée. Elle se base sur les impédances modales du système sous test. Des PCB spécifiques ont été conçus pour valider les équations développées. Une investigation est menée pour définir rigoureusement les impédances modales. Nous avons démontré qu’il y a une divergence entre deux définitions de l’impédance de MC dans la littérature. Ainsi, la conversion de mode (ou rapport Longitudinal Conversion Loss : LCL) a été quantifiée grâce à ces équations. Pour finir, le modèle a été étendu à N-entrées pour représenter un EST de complexité industrielle. Le modèle de l’EST est ensuite associé avec celui de la pince et des câbles travaux réalisés au G2ELAB. Des mesures expérimentales ont été faites pour valider le modèle complet. D’après ces mesures, le courant de MC est impacté par la mise en œuvre des câbles ainsi que celle de l’EST. Il a été montré que la connexion du blindage au plan de masse est le paramètre le plus impactant sur la distribution du courant de MC. / Electronic equipments intended to be integrated in aircrafts are subjected to normative requirements. EMC (Electromagnetic Compatibility) qualification tests became one of the mandatory requirements. This PhD thesis, carried out within the framework of the SIMUCEDO project (SIMulation CEM based on the DO-160 standard), contributes to the modeling of the Bulk Current Injection (BCI) qualification test. Concept, detailed in section 20 in the DO-160 standard, is to generate a noise current via cables using probe injection, then monitor EUT satisfactorily during test. Among the qualification tests, the BCI test is one of the most constraining and time consuming. Thus, its modeling ensures a saving of time, and a better control of the parameters which influence the success of the equipment under test. The modeling of the test was split in two parts : the equipment under test (EUT) on one hand, and the injection probe with the cables on the other hand. This thesis focuses on the EUT modeling. A "gray box" modeling was proposed by associating the "black box" model with the "extensive" model. The gray box is based on the measurement of standard impedances. Its identification is done with a "pi" model. The model, having the advantage of taking into account several configurations of the EUT, has been validated on an analog to digital converter (ADC). Another approach called modal, in function of common mode and differential mode, has been proposed. It takes into account the mode conversion when the EUT is asymmetrical. Specific PCBs were designed to validate the developed equations. An investigation was carried out to rigorously define the modal impedances, in particular the common mode (CM) impedance. We have shown that there is a discrepancy between two definitions of CM impedance in the literature. Furthermore, the mode conversion ratio (or the Longitudinal Conversion Loss : LCL) was quantified using analytical equations based on the modal approach. An N-input model has been extended to include industrial complexity. The EUT model is combined with the clamp and the cables model (made by the G2ELAB laboratory). Experimental measurements have been made to validate the combined model. According to these measurements, the CM current is influenced by the setup of the cables as well as the EUT. It has been shown that the connection of the shield to the ground plane is the most influent parameter on the CM current distribution.

Lors de l'analyse 3D du mouvement humain basée sur des marqueurs cutanés, la position des os ne peut être qu'indirectement estimée. Au cours d'une tâche, les déformations des tissus mous génèrent des déplacements des marqueurs par rapport à l'os : les artefacts de tissus mous (STA), entraînant des effets dévastateurs sur l'estimation de la position. La compensation des STA demeure une question ouverte. L'objectif de cette thèse est de contribuer à la solution de cette question cruciale. La modélisation des STA en utilisant des variables spécifiques mesurables est une condition préalable à son élimination. Un modèle corrigeant les trajectoires individuelles de marqueurs de la cuisse, calibré par des mesures directes des STA, est d'abord présenté. Les STA sont modélisés comme une combinaison linéaire des angles articulaires impliqués. Trois représentations des STA par une série de modes sont proposées : déplacements de marqueurs individuels, transformations géométriques de clusters de marqueurs (MCGT), et variations de forme de l'enveloppe de peau. Le MCGT permet de dissocier les composantes rigides et non rigides. Il a été démontré que seule la composante rigide affecte la cinématique articulaire. Un modèle de cette composante est alors défini pour les clusters cuisse et jambe. Un compromis acceptable entre la correction des STA et le nombre de paramètres a ainsi été obtenu. Les principales applications sont de générer une simulation réaliste des STA ; et surtout, en se concentrant sur la composante rigide, le modèle permet une reconstruction satisfaisante des STA avec moins de paramètres, ce qui facilite son incorporation dans un algorithme d'estimation de la position osseuse / In 3D human movement analysis performed using stereophotogrammetry and skin markers, bone pose can be only indirectly estimated. During a task, soft tissue deformations make the markers move with respect to the underlying bone generating soft tissue artefacts (STA), causing devastating effects on pose estimation and its compensation remains an open issue. The thesis’ aim was to contribute to the solution of this crucial issue. Modelling STA using measurable trial-specific variables is a prerequisite for its removal from marker trajectories. Two STA model architectures are proposed. A thigh marker-level model is first presented. STA was modeled as a linear combination of joint angles involved in the task. The model was calibrated with direct STA measures. The considerable number of model parameters led to defining STA approximations. Three definitions were proposed to represent STA as series of modes : individual marker displacements, marker-cluster geometrical transformations (MCGT), and skin envelope shape variations. Modes were selected using two criteria : modal energy and selecting them a priori. The MCGT allows to select either rigid or non-rigid components. It was also demonstrated that only the rigid component affects joint kinematics. A model of thigh and shank rigid component at cluster-level was then defined. An acceptable trade-off between STA compensation and number of parameters was obtained. These results lead to two main potential applications : generate realistic STAs for simulation