Tracking and understanding human gait is an important step towards improving elderly mobility and safety. This thesis presents a vision-based tracking system that estimates the 3D pose of a wheeled walker user's lower limbs with cameras mounted on the moving walker. The tracker estimates 3D poses from images of the lower limbs in the coronal plane in a dynamic, uncontrolled environment. It employs a probabilistic approach based on particle filtering with three different camera setups: a monocular RGB camera, binocular RGB cameras, and a depth camera. For the RGB cameras, observation likelihoods are designed to compare the colors and gradients of each frame with initial templates that are manually extracted. Two strategies are also investigated for handling appearance change of tracking target: increasing number of templates and using different representations of colors. For the depth camera, two observation likelihoods are developed: the first one works directly in the 3D space, while the second one works in the projected image space. Experiments are conducted to evaluate the performance of the tracking system with different users for all three camera setups. It is demonstrated that the trackers with the RGB cameras produce results with higher error as compared to the depth camera, and the strategies for handling appearance change improve tracking accuracy in general. On the other hand, the tracker with the depth sensor successfully tracks the 3D poses of users over the entire video sequence and is robust against unfavorable conditions such as partial occlusion, missing observations, and deformable tracking target.

In this thesis two different problems in distributed sensor networks are considered. Part I involves optimal quantiser design for decentralised estimation of a two-state hidden Markov model with dual sensors. The notion of optimality for quantiser design is based on minimising the probability of error in estimating the hidden Markov state. Equations for the filter error are derived for the continuous (unquantised) sensor outputs (signals), which are used to benchmark the performance of the quantisers. Minimising the probability of filter error to obtain the quantiser breakpoints is a difficult problem therefore an alternative method is employed. The quantiser breakpoints are obtained by maximising the mutual information between the quantised signals and the hidden Markov state. This method is known to work well for the single sensor case. Cases with independent and correlated noise across the signals are considered. The method is then applied to Markov processes with Gaussian signal noise, and further investigated through simulation studies. Simulations involving both independent and correlated noise across the sensors are performed and a number of interesting new theoretical results are obtained, particularly in the case of correlated noise. In Part II, the focus shifts to the detection of faults in helicopter transmission systems. The aim of the investigation is to determine whether the acoustic signature can be used for fault detection and diagnosis. To investigate this, statistical change detection algorithms are applied to acoustic vibration data obtained from the main rotor gearbox of a Bell 206 helicopter, which is run at high load under test conditions.

Les séries temporelles d'images satellitaires sont une source d'information importante pour le suivi des changements spatio-temporels des surfaces terrestres. En outre, le nombre d’images est en augmentation constante. Pour les exploiter pleinement, des outils dédiés au traitement automatique du contenu informationnel sont développés. Néanmoins ces techniques ne satisfont pas complètement les géographes qui exploitent pourtant, de plus en plus couramment, les données extraites des images dans leurs études afin de prédire le futur. Nous proposons dans cette thèse, une méthodologie générique à base d’un modèle de Markov caché pour l’analyse et la prédiction des changements sur une séquence d’images satellitaires. Cette méthodologie présente deux modules : un module de traitement intégrant les descripteurs et les algorithmes classiquement utilisés en interprétation d'images, et un module d’apprentissage basé sur les modèles de Markov cachés. La performance de notre approche est évaluée par des essais d’interprétations des évènements spatio-temporels effectués sur plusieurs sites d’études. Les résultats obtenus permettront d’analyser et de prédire les changements issus des différentes séries temporelles d’images SPOT et LANDSAT pour l’observation des évènements spatio-temporels telle que l'expansion urbaine et la déforestation. / The time series of satellite images are an important source of information for monitoring spatiotemporal changes of land surfaces. Furthermore, the number of satellite images is increasing constantly, for taking full advantage, tools dedicated to the automatic processing of information content is developed. However these techniques do not completely satisfy the geographers who exploit more currently, the data extracted from the images in their studies to predict the future. In this research we propose a generic methodology based on a hidden Markov model for analyzing and predicting changes in a sequence of satellite images. The methodology that is proposed presents two modules : a processing module which incorporating descriptors and algorithms conventionally used in image interpretation and a learning module based on hidden Markov models. The performance of the approach is evaluated by trials of interpretation of spatiotemporal events conducted in several study sites. Results obtained allow us to analyze and to predict changes from various time series of SPOT and LANDSAT images for observation of spatiotemporal events such as urban development and deforestation.

Made available in DSpace on 2016-06-02T20:06:12Z (GMT). No. of bitstreams: 1 DissDAZ.pdf: 2962567 bytes, checksum: 5c6271a67fae12d6b0160ac8ed9351a2 (MD5) Previous issue date: 2006-02-20 / Universidade Federal de Minas Gerais / (See full text for download) / A crescente necessidade do desenvolvimento de eficientes técnicas computacionais e estatísticas para analisar a profusão de dados biológicos transformaram o modelo Markoviano oculto (HMM), caso particular das redes bayesianas ou probabilísticas, em uma alternativa interessante para analisar sequências de DNA. Uma razão do interesse no HMM é a sua flexibilidade em descrever segmentos heterogêneos da sequência através de uma mesma estrutura de dependência entre as variáveis, supostamente conhecida. No entanto, na maioria dos problemas práticos, a estrutura de dependência não é conhecida e precisa ser também estimada. A maneira mais comum para estimação de estrutra de um HMM é o uso de métodos de seleção de modelos. Outra solução é a utilização de metodologias para estimação da estrutura de uma rede probabilística. Neste trabalho, propomos o HMM de segunda ordem e seus estimadores bayesianos, definimos o fator de Bayes e o DIC para seleção do HMM mais adequado a uma sequência específica, verificamos seus desempenhos e a performance da metodologia proposta por Friedman e Koller (2003) em conjunto de dados simulados e aplicamos estas metodologias em duas sequências de DNA: o intron 7 do gene a - fetoprotein dos cimpanzés e o genoma do parasita Bacteriophage lambda, para o qual o modelo de segunda ordem é mais adequado.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / This work is a study of the relationship between the intrinsic dimension of feature vectors applied to the classification of video signals in order to perform lip reading. In pattern recognition tasks, the extraction of relevant features is crucial for a good performance of the classifiers. The starting point of this work was the reproduction of the work of J.R. Movellan [1], which classifies lips gestures with HMM using only the video signal from the Tulips1 database. The database consists of videos of volunteers’ mouths while they utter the first 4 numerals in English. The original work uses feature vectors of high dimensionality in relation to the size of the database. Consequently, the adjustment of HMM classifiers has become problematic and the maximum accuracy was only 66.67%. Alternative strategies for feature extraction and classification schemes were proposed in order to analyze the influence of the intrinsic dimension in the performance of classifiers. The best solution, in terms of results, achieved an accuracy of approximately 83%. / Este trabalho é um estudo da relação entre a dimensão intrínseca de vetores de características aplicados à classificação de sinais de vídeo no intuito de realizar-se a leitura labial. Nas tarefas de reconhecimento de padrões, a extração de características relevantes é crucial para um bom desempenho dos classificadores. O ponto de partida deste trabalho foi a reprodução do trabalho de J.R. Movellan [1], que realiza a classificação de gestos labiais com HMM na base de dados Tulips1, utilizando somente o sinal de vídeo. A base é composta por vídeos das bocas de voluntários enquanto esses pronunciam os primeiros 4 numerais em inglês. O trabalho original utiliza vetores de características de dimensão muito alta em relação ao tamanho da base. Consequentemente, o ajuste de classificadores HMM se tornou problemático e só se alcançou 66,67% de acurácia. Estratégias de extração de características e esquemas de classificação alternativos foram propostos, a fim de analisar a influência da dimensão intrínseca no desempenho de classificadores. A melhor solução, em termos de resultados, obteve uma acurácia de aproximadamente 83%. / São Cristóvão, SE

Current Ambient Intelligence and Intelligent Environment research focuses on the interpretation of a subject’s behaviour at the activity level by logging the Activity of Daily Living (ADL) such as eating, cooking, etc. In general, the sensors employed (e.g. PIR sensors, contact sensors) provide low resolution information. Meanwhile, the expansion of ubiquitous computing allows researchers to gather additional information from different types of sensor which is possible to improve activity analysis. Based on the previous research about sitting posture detection, this research attempts to further analyses human sitting activity. The aim of this research is to use non-intrusive low cost pressure sensor embedded chair system to recognize a subject’s activity by using their detected postures. There are three steps for this research, the first step is to find a hardware solution for low cost sitting posture detection, second step is to find a suitable strategy of sitting posture detection and the last step is to correlate the time-ordered sitting posture sequences with sitting activity. The author initiated a prototype type of sensing system called IntelliChair for sitting posture detection. Two experiments are proceeded in order to determine the hardware architecture of IntelliChair system. The prototype looks at the sensor selection and integration of various sensor and indicates the best for a low cost, non-intrusive system. Subsequently, this research implements signal process theory to explore the frequency feature of sitting posture, for the purpose of determining a suitable sampling rate for IntelliChair system. For second and third step, ten subjects are recruited for the sitting posture data and sitting activity data collection. The former dataset is collected byasking subjects to perform certain pre-defined sitting postures on IntelliChair and it is used for posture recognition experiment. The latter dataset is collected by asking the subjects to perform their normal sitting activity routine on IntelliChair for four hours, and the dataset is used for activity modelling and recognition experiment. For the posture recognition experiment, two Support Vector Machine (SVM) based classifiers are trained (one for spine postures and the other one for leg postures), and their performance evaluated. Hidden Markov Model is utilized for sitting activity modelling and recognition in order to establish the selected sitting activities from sitting posture sequences.2. After experimenting with possible sensors, Force Sensing Resistor (FSR) is selected as the pressure sensing unit for IntelliChair. Eight FSRs are mounted on the seat and back of a chair to gather haptic (i.e., touch-based) posture information. Furthermore, the research explores the possibility of using alternative non-intrusive sensing technology (i.e. vision based Kinect Sensor from Microsoft) and find out the Kinect sensor is not reliable for sitting posture detection due to the joint drifting problem. A suitable sampling rate for IntelliChair is determined according to the experiment result which is 6 Hz. The posture classification performance shows that the SVM based classifier is robust to “familiar” subject data (accuracy is 99.8% with spine postures and 99.9% with leg postures). When dealing with “unfamiliar” subject data, the accuracy is 80.7% for spine posture classification and 42.3% for leg posture classification. The result of activity recognition achieves 41.27% accuracy among four selected activities (i.e. relax, play game, working with PC and watching video). The result of this thesis shows that different individual body characteristics and sitting habits influence both sitting posture and sitting activity recognition. In this case, it suggests that IntelliChair is suitable for individual usage but a training stage is required.

Cette thèse est consacrée à l'évaluation de la fiabilité d’une caravane de véhicules intelligents avec les principaux objectifs de fournir une méthode pour évaluer et améliorer la sécurité des ITS. Après examen de l'évolution des ITS et DSRC, la fiabilité/qualité de service (QoS) de transmission de bout-en-bout des communications courtes est analysée. Ensuite, les effets de la transmission et QoS sur le système de commande PID et flou sont étudiés respectivement.D'abord, Les méthodes de protection des données utilisées dans le protocole Bluetooth sont analysées en termes de fiabilité. Un modèle de performance du nœud est construit pour les différentes distributions de temps de service du nœud. Ainsi, chaque nœud dans le réseau peut surveiller directement la qualité de service du nœud à tout moment. Basé sur ce modèle, un protocole de routage ad hoc (NPDSR) est proposé. Ce protocole présente les avantages de maintenir la qualité de service requise de la communication, et des économies de routage.Deuxièmement, les effets sur les systèmes de contrôle, en termes de retards et de pertes, sont analysés en utilisant la méthode mathématique et statistique. Des fonctions descriptives pour les contrôleurs flous de types Mamdani et T-S sont fournies, et la stabilité du contrôleur multi-flou avec un retard de communication est analysée à l'aide des lieux de Nyquist.Enfin, en perspective, l’utilisation possible de NS-2 sur le modèle de simulation et de vérification est discutée, ce qui s’avère intéressant pour simuler des protocoles différents. En outre, il peut être connecté avec Simulink permettant une meilleure analyse de performances des systèmes de contrôle en réseau / This PhD thesis is dedicated to the reliability assessment for intelligent vehicles platoon with the main aim of providing an approach to evaluate and enhance the safety of ITS. After a review of developments of ITS and DSRC, the reliability/QoS of end-to-end transmission of short range communication is analyzed. Then, the effects of transmission QoS on PID sampling control system and fuzzy control system are studied respectively.At first, The data protection methods used in Bluetooth communication protocol are analyzed for the data reliability. A node performance model is built, suitable for different distributions of service time of the node. Hence, a node in the network can monitor the QoS of the node by itself at any time. Based on this model, a QoS-aware ad hoc routing protocol (NPDSR) is proposed. This protocol has the advantages in maintaining requested QoS of communication, and it has less routing overhead.Secondly, for the effects on the control systems, time delay and loss are both discussed using mathematical and statistical simulation methods. Describing functions for both Mamdani and T-S type fuzzy controllers are provided. The stability of multi-fuzzy controller with communication delay is analyzed using Nyquist stability plots.Finally, in the perspective, the possible use of NS-2 on the model simulation and verification has been discussed, which is powerful in simulating different behavior of protocols, routing protocol comparisons. Also, it can be connected with Simulink to enhance performance the simulation of network control systems

Cette thèse est dédiée à la détection de l'ombre, de la végétation et des bâtiments à partir d'une unique image optique très haute résolution. La première partie présente une nouvelle méthode pour détecter simultanément les ombres et la végétation : plusieurs indices d'ombre et de végétation sont comparés puis fusionnés grâce à la théorie de l'évidence de Dempster-Shafer afin d'obtenir une segmentation en trois classes : “ombre”, “végétation” et “autre”. Comme la fusion est une méthode pixellique, elle est incorporée dans un contexte markovien pour régulariser la segmentation. Dans la deuxième partie, une nouvelle technique de segmentation d'images par croissance de région est proposée. L'image est tout d'abord sur-segmentée en régions homogènes afin de remplacer la structure rigide de la grille de pixels. Une classification-fusion itérative est ensuite appliquée sur ces régions. À chaque itération, les régions sont classées en utilisant une segmentation markovienne, puis regroupées entre elles en fonction de la position des ombres, de leur classe, et de la rectangularité de la forme fusionnée. Les bâtiments sont estimés à partir de la classification finale comme étant les rectangles d'emprise minimale. Ces deux algorithmes ont été validés sur plusieurs images de télédétection et ont permis de démontrer leur efficacité. / This PhD thesis is devoted to the detection of shadows, vegetation and buildings from single high resolution optical remote sensing images. The first part introduces a new method for simultaneously detecting shadows and vegetation. Several shadow and vegetation indices were investigated and merged using the Dempster-Shafer evidence theory so as to obtain a segmentation map with three classes : “shadow”, “vegetation” and “other”. However, the performance of the fusion is sensitive to noise since it processes at a pixel-level. A Markov random field (MRF) is thus integrated to model spatial information within the image. In the second part, a novel region growing segmentation technique is proposed. The image is oversegmented into smaller homogeneous regions which replace the rigid structure of the pixel grid. An iterative region classification-merging is then applied over these regions. At each iteration, regions are classified using a MRF-based image segmentation, then, according to the position of shadows, regions having the same class are merged to produce shapes appropriate to rectangles. The final buildings are estimated using the recursive minimum bounding rectangle method from the final classification. These two algorithms have been validated on a variety of image datasets and demonstrate their efficiency.

Parmi les manifestations de la maltraitance de l’enfant, problème de santé publique, le traumatisme crânien infligé de l’enfant est une des premières causes de mortalité par maltraitance. L’objectif de cette thèse était d’évaluer l’opportunité d’un dépistage de la maltraitance physique de l’enfant. Le premier obstacle mis en évidence était l’imprécision et le manque de standardisation de la définition de la maltraitance physique. Face à cet obstacle,nous nous sommes interrogés sur la perception qu’avaient les professionnels de santé et les parents des urgences pédiatriques du CHU de Bordeaux de la violence physique des parents vis-à-vis de leur enfant. Les professionnels étaient plus tolérants vis-à-vis de ces violences.Une même situation était évaluée différemment selon le professionnel. Nous avons illustré la variabilité de perception d’une même situation lorsque l’enfant ne présente pas encore de conséquences graves de maltraitance à travers un cas clinique de traumatisme crânien infligé,ainsi que les craintes d’une stigmatisation des parents directement en lien avec la difficulté à identifier précocement ces situations et les effets négatifs que peut avoir la prise en charge.Enfin, nous avons simulé par un modèle de Markov, en tenant compte des incertitudes identifiées, l’impact d’un programme de prévention primaire et d’un programme de dépistage du traumatisme crânien infligé. Ce modèle a permis de confirmer l’intérêt d’un programme de prévention primaire. Il a également montré l’importance de déterminer les effets possibles d’une stigmatisation à tort des parents dans le cadre du dépistage pour s’assurer qu’un tel programme ne serait pas néfaste. / Among child abuse, which is a major public health issue, Pediatric Abusive Head Trauma is one of the major causes of death in abused children. The aim of this thesis was to assess the opportunity of screening for physical child abuse. The first identified obstacle was the lack of knowledge about the beginning of physical child abuse, and the vagueness and lack ofstandardization of its definition. Therefore, we conducted a study to assess the perception of physical violence by parents toward their children, by parents and professionals in the emergency department of the university hospital of Bordeaux. The professionals were more tolerant than parents and the perception of a same situation could vary according to the professional. We have illustrated this variation with a clinical case of Pediatric Abusive Head Trauma, when the child had not already serious consequences of child abuse. Similarly, the difficulty of early identification of abused children was responsible for the fear of mother’sstigmatization. And we discussed the impact of management of the children and their mother.Finally, we evaluated the impact of a primary prevention program and screening program ofPediatric Abusive Head Trauma using a Markov model considering identified uncertainties.The simulation confirmed the potential benefits of primary prevention program documentedthe huge uncertainty regarding benefits associated with screening of Pediatric Abusive HeadTrauma. Future research should in particular focus on describing the effects of wrong lyidentifying parents as abusers.

The vital signs monitoring data of an infant receiving intensive care are a rich source of information about its health condition. One major concern about the state of health of such patients is the onset of neonatal sepsis, a life-threatening bloodstream infection. As early signs are subtle and current diagnosis procedures involve slow laboratory testing, sepsis detection based on the monitored physiological dynamics is a clinically significant task. This challenging problem can be thoroughly modelled as real-time inference within a machine learning framework. In this thesis, we develop probabilistic dynamical models centred around the goal of providing useful predictions about the onset of neonatal sepsis. This research is characterised by the careful incorporation of domain knowledge for the purpose of extracting the infant’s true physiology from the monitoring data. We make two main contributions. The first one is the formulation of sepsis detection as learning and inference in an Auto-Regressive Hidden Markov Model (AR-HMM). The model investigates the extent to which physiological events observed in the patient’s monitoring traces could be used for the early detection of neonatal sepsis. In addition, the proposed approach involves exact marginalisation over missing data at inference time. When applying the ARHMM on a real-world dataset, we found that it can produce effective predictions about the onset of sepsis. Second, both sepsis and clinical event detection are formulated as learning and inference in a Hierarchical Switching Linear Dynamical System (HSLDS). The HSLDS models dynamical systems where complex interactions between modes of operation can be represented as a twolevel hidden discrete hierarchical structure. For neonatal condition monitoring, the lower layer models clinical events and is controlled by upper layer variables with semantics sepsis/nonsepsis. The model parameterisation and estimation procedures are adapted to the specifics of physiological monitoring data. We demonstrate that the performance of the HSLDS for the detection of sepsis is not statistically different from the AR-HMM, despite the fact that the latter model is given “ground truth” annotations of the patient’s physiology.