Observa-se nos tempos atuais um crescente interesse e demanda por novas tecnologias de sensoriamento e interação. A monitoração, com o objetivo de reconhecimento de movimentos humanos, permite oferecer serviços personalizados em diferentes áreas, dentre elas a área de cuidados médicos. Essa monitoração pode ser realizada por meio de diferentes técnicas como o uso de câmeras de vídeo, instrumentação do ambiente onde o indivíduo habita, ou pelo uso de dispositivos pessoais acoplados ao corpo. Os dispositivos acoplados ao corpo apresentam vantagens como baixo custo, uso confortável, além de muitas vezes serem despercebidos pelo usuário, diminuindo a sensação de invasão de privacidade durante a monitoração. Além disso, o dispositivo sensor pode ser facilmente acoplado ao corpo pelo próprio usuário, tornando o seu uso efetivo. Deste modo, este trabalho apresenta o desenvolvimento de um sistema que emprega técnicas de inteligência computacional e um acelerômetro facilmente acoplado ao punho do usuário para efetuar, de maneira confortável e não invasiva, o reconhecimento de movimentos básicos da rotina de uma pessoa. Aplicando máquinas de vetores de suporte para classificar os sinais e a razão discriminante de Fisher para efetuar a seleção das características mais significativas, o sistema apresentou uma taxa de sucesso em torno de 93% no reconhecimento de movimentos básicos efetuados por indivíduos monitorados. O sistema apresenta potencialidade para ser integrado a um hardware embarcado de baixo custo, responsável pelo gerenciamento da aquisição dos dados e pelo encaminhamento das informações a um sistema de monitoramento ou armazenamento. As informações providas por este sistema podem ser destinadas à promoção da saúde e bem estar do indivíduo, bem como utilizadas em diagnósticos ou monitoramento remoto de pacientes em um ambiente de vida assistida. / Nowadays it is observed a growing interest and demand for new sensing technologies and interaction. Monitoring with the objective of recognizing human movements, allows us to offer personalized services in different areas, among them healthcare. This monitoring can be performed through the use of different techniques such as the use of video cameras, living environment instrumentation, or the use of personal devices attached to the body, also known as wearable devices. These wearable devices have some advantages such as low cost, comfortable to use, and are often unnoticed by the user, reducing the feeling of privacy invasion during the monitoring. In addition, the sensing device can be easily attached to the body by the user itself, making its use effective. Thus, this work presents the development of a system that uses computational intelligence techniques and an accelerometer which is easily attached to the users wrist to perform, in a comfortable and non-invasive manner, the recognition of basic movements of a persons routine. By applying support vector machines to classify the signals and Fishers discriminant ratio to select the most significant features, the system has shown a success rate of 93% in the recognition of basic movements performed by monitored individuals. The system has the potential to be integrated into a low-cost embedded hardware, which is responsible for managing the data acquisition and routing the movement data to a remote monitoring system or storage. The information provided by the system can be designed to promote the health and wellness of the individual, as well used in diagnostics or remote patient monitoring in an ambient assisted living (AAL).

Observa-se nos tempos atuais um crescente interesse e demanda por novas tecnologias de sensoriamento e interação. A monitoração, com o objetivo de reconhecimento de movimentos humanos, permite oferecer serviços personalizados em diferentes áreas, dentre elas a área de cuidados médicos. Essa monitoração pode ser realizada por meio de diferentes técnicas como o uso de câmeras de vídeo, instrumentação do ambiente onde o indivíduo habita, ou pelo uso de dispositivos pessoais acoplados ao corpo. Os dispositivos acoplados ao corpo apresentam vantagens como baixo custo, uso confortável, além de muitas vezes serem despercebidos pelo usuário, diminuindo a sensação de invasão de privacidade durante a monitoração. Além disso, o dispositivo sensor pode ser facilmente acoplado ao corpo pelo próprio usuário, tornando o seu uso efetivo. Deste modo, este trabalho apresenta o desenvolvimento de um sistema que emprega técnicas de inteligência computacional e um acelerômetro facilmente acoplado ao punho do usuário para efetuar, de maneira confortável e não invasiva, o reconhecimento de movimentos básicos da rotina de uma pessoa. Aplicando máquinas de vetores de suporte para classificar os sinais e a razão discriminante de Fisher para efetuar a seleção das características mais significativas, o sistema apresentou uma taxa de sucesso em torno de 93% no reconhecimento de movimentos básicos efetuados por indivíduos monitorados. O sistema apresenta potencialidade para ser integrado a um hardware embarcado de baixo custo, responsável pelo gerenciamento da aquisição dos dados e pelo encaminhamento das informações a um sistema de monitoramento ou armazenamento. As informações providas por este sistema podem ser destinadas à promoção da saúde e bem estar do indivíduo, bem como utilizadas em diagnósticos ou monitoramento remoto de pacientes em um ambiente de vida assistida. / Nowadays it is observed a growing interest and demand for new sensing technologies and interaction. Monitoring with the objective of recognizing human movements, allows us to offer personalized services in different areas, among them healthcare. This monitoring can be performed through the use of different techniques such as the use of video cameras, living environment instrumentation, or the use of personal devices attached to the body, also known as wearable devices. These wearable devices have some advantages such as low cost, comfortable to use, and are often unnoticed by the user, reducing the feeling of privacy invasion during the monitoring. In addition, the sensing device can be easily attached to the body by the user itself, making its use effective. Thus, this work presents the development of a system that uses computational intelligence techniques and an accelerometer which is easily attached to the users wrist to perform, in a comfortable and non-invasive manner, the recognition of basic movements of a persons routine. By applying support vector machines to classify the signals and Fishers discriminant ratio to select the most significant features, the system has shown a success rate of 93% in the recognition of basic movements performed by monitored individuals. The system has the potential to be integrated into a low-cost embedded hardware, which is responsible for managing the data acquisition and routing the movement data to a remote monitoring system or storage. The information provided by the system can be designed to promote the health and wellness of the individual, as well used in diagnostics or remote patient monitoring in an ambient assisted living (AAL).

There are many amputees around the world who have lost a limb through conflict, disease or an accident. Upper-limb prostheses controlled using surface Electromyography (sEMG) offer a solution to help the amputees; however, their functionality is limited by the small number of movements they can perform and their slow reaction times. Pattern recognition (PR)-based EMG control has been proposed to improve the functional performance of prostheses. It is a very promising approach, offering intuitive control, fast reaction times and the ability to control a large number of degrees of freedom (DOF). However, prostheses controlled with PR systems are not available for everyday use by amputees, because there are many major challenges and practical problems that need to be addressed before clinical implementation is possible. These include lack of individual finger control, an impractically large number of EMG electrodes, and the lack of deployment protocols for EMG electrodes site selection and movement optimisation. Moreover, the inability of PR systems to handle multiple forces is a further practical problem that needs to be addressed. The main aim of this project is to investigate the research challenges mentioned above via non-invasive EMG signal acquisition, and to propose practical solutions to help amputees. In a series of experiments, the PR systems presented here were tested with EMG signals acquired from seven transradial amputees, which is unique to this project. Previous studies have been conducted using non-amputees. In this work, the challenges described are addressed and a new protocol is proposed that delivers a fast clinical deployment of multi-functional upper limb prostheses controlled by PR systems. Controlling finger movement is a step towards the restoration of lost human capabilities, and is psychologically important, as well as physically. A central thread running through this work is the assertion that no two amputees are the same, each suffering different injuries and retaining differing nerve and muscle structures. This work is very much about individualised healthcare, and aims to provide the best possible solution for each affected individual on a case-by-case basis. Therefore, the approach has been to optimise the solution (in terms of function and reliability) for each individual, as opposed to developing a generic solution, where performance is optimised against a test population. This work is unique, in that it contributes to improving the quality of life for each individual amputee by optimising function and reliability. The main four contributions of the thesis are as follows: 1- Individual finger control was achieved with high accuracy for a large number of finger movements, using six optimally placed sEMG channels. This was validated on EMG signals for ten non-amputee and six amputee subjects. Thumb movements were classified successfully with high accuracy for the first time. The outcome of this investigation will help to add more movements to the prosthesis, and reduce hardware and computational complexity. 2- A new subject-specific protocol for sEMG site selection and reliable movement subset optimisation, based on the amputee’s needs, has been proposed and validated on seven amputees. This protocol will help clinicians to perform an efficient and fast deployment of prostheses, by finding the optimal number and locations of EMG channels. It will also find a reliable subset of movements that can be achieved with high performance. 3- The relationship between the force of contraction and the statistics of EMG signals has been investigated, utilising an experimental design where visual feedback from a Myoelectric Control Interface (MCI) helped the participants to produce the correct level of force. Kurtosis values were found to decrease monotonically when the contraction level increased, thus indicating that kurtosis can be used to distinguish different forces of contractions. 4- The real practical problem of the degradation of classification performance as a result of the variation of force levels during daily use of the prosthesis has been investigated, and solved by proposing a training approach and the use of a robust feature extraction method, based on the spectrum. The recommendations of this investigation improve the practical robustness of prostheses controlled with PR systems and progress a step further towards clinical implementation and improving the quality of life of amputees. The project showed that PR systems achieved a reliable performance for a large number of amputees, taking into account real life issues such as individual finger control for high dexterity, the effect of force level variation, and optimisation of the movements and EMG channels for each individual amputee. The findings of this thesis showed that the PR systems need to be appropriately tuned before usage, such as training with multiple forces to help to reduce the effect of force variation, aiming to improve practical robustness, and also finding the optimal EMG channel for each amputee, to improve the PR system’s performance. The outcome of this research enables the implementation of PR systems in real prostheses that can be used by amputees.

Cette thèse est une collaboration entre le LAboratoire des Sciences et Matériaux pour l’Électronique et d’Automatique (LASMEA) de Clermont-Ferrand et le Laboratoire Vision et Ingénierie des Contenus (LVIC) du CEA LIST à Saclay. La première moitié de la thèse a été accomplie au sein de l’équipe ComSee (1) du LASMEA et la deuxième au LVIC. L’objectif de ces travaux est de concevoir un système de vidéo-assistance temps réel pour la détection d’évènements dans des scènes possiblement denses.La vidéosurveillance intelligente de scènes denses telles que des foules est particulièrement difficile, principalement à cause de leur complexité et de la grande quantité de données à traiter simultanément. Le but de cette thèse consiste à élaborer une méthode de détection d’évènements rares dans de telles scènes, observées depuis une caméra fixe. La méthode en question s’appuie sur l’analyse automatique de mouvement et ne nécessite aucune information à priori. Les mouvements nominaux sont déterminés grâce à un apprentissage statistique non supervisé. Les plus fréquemment observés sont considérés comme des évènements normaux. Une phase de classification permet ensuite de détecter les mouvements déviant trop du modèle statistique, pour les considérer comme anormaux. Cette approche est particulièrement adaptée aux lieux de déplacements structurés, tels que des scènes de couloirs ou de carrefours routiers. Aucune étape de calibration, de segmentation de l’image, de détection d’objets ou de suivi n’est nécessaire. Contrairement aux analyses de trajectoires d’objets suivis, le coût calculatoire de notre méthode est invariante au nombre de cibles présentes en même temps et fonctionne en temps réel. Notre système s’appuie sur une classification locale du mouvement de la scène, sans calibration préalable. Dans un premier temps, une caractérisation du mouvement est réalisée, soit par des méthodes classiques de flot optique, soit par des descripteurs spatio-temporels. Ainsi, nous proposons un nouveau descripteur spatio-temporel fondé sur la recherche d’une relation linéaire entre les gradients spatiaux et les gradients temporels en des zones où le mouvement est supposé uniforme. Tout comme les algorithmes de flot optique, ce descripteur s’appuie sur la contrainte d’illumination constante.Cependant en prenant en compte un voisinage temporel plus important, il permet une caractérisation du mouvement plus lisse et plus robuste au bruit. De plus, sa faible complexité calculatoire est bien adaptée aux applications temps réel. Nous proposons ensuite d’étudier différentes méthodes de classification : La première, statique, dans un traitement image par image, s’appuie sur une estimation bayésienne de la caractérisation du mouvement au travers d’une approche basée sur les fenêtres de Parzen. Cette nouvelle méthode est une variante parcimonieuse des fenêtres de Parzen. Nous montrons que cette approche est algorithmiquement efficace pour approximer de manière compacte et précise les densités de probabilité. La seconde méthode, basée sur les réseaux bayésiens, permet de modéliser la dynamique du mouvement. Au lieu de considérer ce dernier image par image, des séquences de mouvements sont analysées au travers de chaînes de Markov Cachées. Ajouté à cela, une autre contribution de ce manuscrit est de prendre en compte la modélisation du voisinage d’un bloc afin d’ajouter une cohérence spatiale à la propagation du mouvement. Ceci est réalisé par le biais de couplages de chaînes de Markov cachées.Ces différentes approches statistiques ont été évaluées sur des données synthétiques ainsi qu’en situations réelles, aussi bien pour la surveillance du trafic routier que pour la surveillance de foule.Cette phase d’évaluation permet de donner des premières conclusions encourageantes quant à la faisabilité de la vidéosurveillance intelligente d’espaces possiblement denses. / The automatic analysis of crowded areas in video sequences is particularly difficult because ofthe large amount of information to be processed simultaneously and the complexity of the scenes. We propose in this thesis a method for detecting abnormal events in possibly dense scenes observed from a static camera. The approach is based on the automatic classification of motion requiring no prior information. Motion patterns are encoded in an unsupervised learning framework in order to generate a statistical model of frequently observed (aka. normal) events. Then at the detection stage, motion patterns that deviate from the model are classified as unexpected events. The method is particularly adapted to scenes with structured movement with directional flow of objects or people such as corridors, roads, intersections. No camera calibration is needed, nor image segmentation, object detection and tracking. In contrast to approaches that rely on trajectory analysis of tracked objects, our method is independent of the number of targets and runs in real-time. Our system relies on a local classification of global scene movement. The local analysis is done on each blocks of a regular grid. We first introduce a new spatio-temporal local descriptor to characterize the movement efficiently. Assuming a locally uniform motion of space-time blocks of the image, our approach consists in determining whether there is a linear relationship between spatial gradients and temporal gradients. This spatio-temporal descriptor holds the Illumination constancy constraint like optical flow techniques, but it allows taking into account the spatial neighborhood and a temporal window by giving a smooth characterization of the motion, which makes it more robust to noise. In addition, its low computational complexity is suitable for real-time applications. Secondly, we present two different classification frameworks : The first approach is a static (frame by frame) classification approach based on a Bayesian characterization of the motion by using an approximation of the Parzen windowing method or Kernel Density Estimation (KDE) to model the probability density function of motion patterns.This new method is the sparse variant of the KDE (SKDE). We show that the SKDE is a very efficient algorithm giving compact representations and good approximations of the density functions. The second approach, based on Bayesian Networks, models the dynamics of the movement. Instead of considering motion patterns in each block independently, temporal sequences of motion patterns are learned by using Hidden Markov Models (HMM). The second proposed improvement consists in modeling the movement in one block by taking into account the observed motion in adjacent blocks. This is performed by the coupled HMM method. Evaluations were conducted to highlight the classification performance of the proposed methods,on both synthetic data and very challenging real video sequences captured by video surveillance cameras.These evaluations allow us to give first conclusions concerning automatic analyses of possibly crowded area.