Rozprawy doktorskie na temat „TRACKING HUMAN BODY”

With the recent developments in Computer Vision and Pattern Recognition, surveillance applications are equipped with the capabilities of event/activity understanding and interpretation which usually require recognizing humans in real world scenes. Real world scenes such as airports, streets and train stations are complex because they involve many people, complicated occlusions and cluttered backgrounds. Although complex real world scenes exist, human detectors have the capability to locate pedestrians accurately even in complex scenes and visual trackers have the capability to track targets in cluttered environments. The integration of visual object detection and tracking, which are the fundamental features of available surveillance applications, is one of the solutions for multi-human tracking problem in crowded scenes which is studied in this thesis. In this thesis, human body part detectors, which are capable of detecting human heads and human upper body parts, are trained with Support Vector Machines (SVM) by using Histogram of Oriented Gradients (HOG), which is one of the state-of-the-art descriptor for human detection. The training process is elaborated by investigating the effects of the parameters of the HOG descriptor. The human heads and upper body parts are searched in the region of interests (ROI) computed by detecting motion. In addition, these human body part detectors are integrated with a multi-human tracker which solves the data association problem with the Multi Scan Markov Chain Monte Carlo Data Association (MCMCDA) algorithm. Associated measurements of human upper body part locations are used for state correction for each track. State estimation is done through Kalman Filter. The performance of detectors are evaluated using MIT Pedestrian dataset and INRIA Human dataset.

Wireless capsule endoscopy (WCE) offers a patient-friendly, non-invasive and painless investigation of the entire small intestine, where other conventional wired endoscopic instruments can barely reach. As a critical component of the capsule endoscopic examination, physicians need to know the precise position of the endoscopic capsule in order to identify the position of intestinal disease after it is detected by the video source. To define the position of the endoscopic capsule, we need to have a map of inside the human body. However, since the shape of the small intestine is extremely complex and the RF signal propagates differently in the non-homogeneous body tissues, accurate mapping and localization inside small intestine is very challenging. In this dissertation, we present an in-body simultaneous localization and mapping technique (Body-SLAM) to enhance the positioning accuracy of the WCE inside the small intestine and reconstruct the trajectory the capsule has traveled. In this way, the positions of the intestinal diseases can be accurately located on the map of inside human body, therefore, facilitates the following up therapeutic operations. The proposed approach takes advantage of data fusion from two sources that come with the WCE: image sequences captured by the WCE's embedded camera and the RF signal emitted by the capsule. This approach estimates the speed and orientation of the endoscopic capsule by analyzing displacements of feature points between consecutive images. Then, it integrates this motion information with the RF measurements by employing a Kalman filter to smooth the localization results and generate the route that the WCE has traveled. The performance of the proposed motion tracking algorithm is validated using empirical data from the patients and this motion model is later imported into a virtual testbed to test the performance of the alternative Body-SLAM algorithms. Experimental results show that the proposed Body-SLAM technique is able to provide accurate tracking of the WCE with average error of less than 2.3cm.

Geometric reconstruction of dynamic objects is a fundamental task of computer vision and graphics, and modeling human body of high fidelity is considered to be a core of this problem. Traditional human shape and motion capture techniques require an array of surrounding cameras or subjects wear reflective markers, resulting in a limitation of working space and portability. In this dissertation, a complete process is designed from geometric modeling detailed 3D human full body and capturing shape dynamics over time using a flexible setup to guiding clothes/person re-targeting with such data-driven models. As the mechanical movement of human body can be considered as an articulate motion, which is easy to guide the skin animation but has difficulties in the reverse process to find parameters from images without manual intervention, we present a novel parametric model, GMM-BlendSCAPE, jointly taking both linear skinning model and the prior art of BlendSCAPE (Blend Shape Completion and Animation for PEople) into consideration and develop a Gaussian Mixture Model (GMM) to infer both body shape and pose from incomplete observations. We show the increased accuracy of joints and skin surface estimation using our model compared to the skeleton based motion tracking. To model the detailed body, we start with capturing high-quality partial 3D scans by using a single-view commercial depth camera. Based on GMM-BlendSCAPE, we can then reconstruct multiple complete static models of large pose difference via our novel non-rigid registration algorithm. With vertex correspondences established, these models can be further converted into a personalized drivable template and used for robust pose tracking in a similar GMM framework. Moreover, we design a general purpose real-time non-rigid deformation algorithm to accelerate this registration. Last but not least, we demonstrate a novel virtual clothes try-on application based on our personalized model utilizing both image and depth cues to synthesize and re-target clothes for single-view videos of different people.

Les robots sont des agents artificiels qui peuvent agir dans le monde des humains grâce aux capacités de perception. Dans un contexte d'interaction homme-robot, les humains et les robots partagent le même espace de communication. En effet, les robots compagnons sont censés communiquer avec les humains d'une manière naturelle et intuitive: l'une des façons les plus naturelles est basée sur les gestes et les mouvements réactifs du corps. Pour rendre cette interaction la plus conviviale possible, un robot compagnon doit, donc, être doté d'une ou plusieurs capacités lui permettant de percevoir, de reconnaître et de réagir aux gestes humains. Cette thèse a été focalisée sur la conception et le développement d'un système de reconnaissance gestuelle dans un contexte d'interaction homme-robot. Ce système comprend un algorithme de suivi permettant de connaître la position du corps lors des mouvements et un module de niveau supérieur qui reconnaît les gestes effectués par des utilisateurs humains. De nouvelles contributions ont été apportées dans les deux sujets. Tout d'abord, une nouvelle approche est proposée pour le suivi visuel des membres du haut du corps. L'analyse du mouvement du corps humain est difficile, en raison du nombre important de degrés de liberté de l'objet articulé qui modélise la partie supérieure du corps. Pour contourner la complexité de calcul, chaque membre est suivi avec un filtre particulaire à recuit simulé et les différents filtres interagissent grâce à la propagation de croyance. Le corps humain en 3D est ainsi qualifié comme un modèle graphique dans lequel les relations entre les parties du corps sont représentées par des distributions de probabilité conditionnelles. Le problème d'estimation de la pose est donc formulé comme une inférence probabiliste sur un modèle graphique, où les variables aléatoires correspondent aux paramètres des membres individuels (position et orientation) et les messages de propagation de croyance assurent la cohérence entre les membres. Deuxièmement, nous proposons un cadre permettant la détection et la reconnaissance des gestes emblématiques. La question la plus difficile dans la reconnaissance des gestes est de trouver de bonnes caractéristiques avec un pouvoir discriminant (faire la distinction entre différents gestes) et une bonne robustesse à la variabilité intrinsèque des gestes (le contexte dans lequel les gestes sont exprimés, la morphologie de la personne, le point de vue, etc). Dans ce travail, nous proposons un nouveau modèle de normalisation de la cinématique du bras reflétant à la fois l'activité musculaire et l'apparence du bras quand un geste est effectué. Les signaux obtenus sont d'abord segmentés et ensuite analysés par deux techniques d'apprentissage : les chaînes de Markov cachées et les Support Vector Machine. Les deux méthodes sont comparées dans une tâche de reconnaissance de 5 classes de gestes emblématiques. Les deux systèmes présentent de bonnes performances avec une base de données de formation minimaliste quels que soient l'anthropométrie, le sexe, l'âge ou la pose de l'acteur par rapport au système de détection. Le travail présenté ici a été réalisé dans le cadre d'une thèse de doctorat en co-tutelle entre l'Université "Pierre et Marie Curie" (ISIR laboratoire, Paris) et l'Université de Gênes (IIT - Tera département) et a été labelisée par l'Université Franco-Italienne.

Les robots sont des agents artificiels qui peuvent agir dans le monde des humains grâce aux capacités de perception. Dans un contexte d’interaction homme-robot, les humains et les robots partagent le même espace de communication. En effet, les robots compagnons sont censés communiquer avec les humains d’une manière naturelle et intuitive: l’une des façons les plus naturelles est basée sur les gestes et les mouvements réactifs du corps. Pour rendre cette interaction la plus conviviale possible, un robot compagnon doit, donc, être doté d’une ou plusieurs capacités lui permettant de percevoir, de reconnaître et de réagir aux gestes humains. Cette thèse a été focalisée sur la conception et le développement d’un système de reconnaissance gestuelle dans un contexte d’interaction homme-robot. Ce système comprend un algorithme de suivi permettant de connaître la position du corps lors des mouvements et un module de niveau supérieur qui reconnaît les gestes effectués par des utilisateurs humains. De nouvelles contributions ont été apportées dans les deux sujets. Tout d’abord, une nouvelle approche est proposée pour le suivi visuel des membres du haut du corps. L’analyse du mouvement du corps humain est difficile, en raison du nombre important de degrés de liberté de l’objet articulé qui modélise la partie supérieure du corps. Pour contourner la complexité de calcul, chaque membre est suivi avec un filtre particulaire à recuit simulé et les différents filtres interagissent grâce à la propagation de croyance. Le corps humain en 3D est ainsi qualifié comme un modèle graphique dans lequel les relations entre les parties du corps sont représentées par des distributions de probabilité conditionnelles. Le problème d’estimation de la pose est donc formulé comme une inférence probabiliste sur un modèle graphique, où les variables aléatoires correspondent aux paramètres des membres individuels (position et orientation) et les messages de propagation de croyance assurent la cohérence entre les membres. Deuxièmement, nous proposons un cadre permettant la détection et la reconnais- sance des gestes emblématiques. La question la plus difficile dans la reconnaissance des gestes est de trouver de bonnes caractéristiques avec un pouvoir discriminant (faire la distinction entre différents gestes) et une bonne robustesse à la variabilité intrinsèque des gestes (le contexte dans lequel les gestes sont exprimés, la morpholo- gie de la personne, le point de vue, etc). Dans ce travail, nous proposons un nouveau modèle de normalisation de la cinématique du bras reflétant à la fois l’activité mus- culaire et l’apparence du bras quand un geste est effectué. Les signaux obtenus sont d’abord segmentés et ensuite analysés par deux techniques d’apprentissage : les chaînes de Markov cachées et les Support Vector Machine. Les deux méthodes sont comparées dans une tâche de reconnaissance de 5 classes de gestes emblématiques. Les deux systèmes présentent de bonnes performances avec une base de données de formation minimaliste quels que soient l’anthropométrie, le sexe, l’âge ou la pose de l’acteur par rapport au système de détection. Le travail présenté ici a été réalisé dans le cadre d’une thèse de doctorat en co-tutelle entre l’Université “Pierre et Marie Curie” (ISIR laboratoire, Paris) et l’Université de Gênes (IIT - Tera département) et a été labelisée par l’Université Franco-Italienne
Robots are artificial agents that can act in humans’ world thanks to perception, action and reasoning capacities. In particular, robots companion are designed to share with humans the same physical and communication spaces in performing daily life collaborative tasks and aids. In such a context, interactions between humans and robots are expected to be as natural and as intuitive as possible. One of the most natural ways is based on gestures and reactive body motions. To make this friendly interaction possible, a robot companion has to be endowed with one or more capabilities allowing him to perceive, to recognize and to react to human gestures. This PhD thesis has been focused on the design and the development of a gesture recognition system that can be exploited in a human-robot interaction context. This system includes (1) a limbs-tracking algorithm that determines human body position during movements and (2) a higher-level module that recognizes gestures performed by human users. New contributions were made in both topics. First, a new approach is proposed for visual tracking of upper-body limbs. Analysing human body motion is challenging, due to the important number of degrees of freedom of the articulated object modelling the upper body. To circumvent the computational complexity, each limb is tracked with an Annealed Particle Filter and the different filters interact through Belief Propagation. 3D human body is described as a graphical model in which the relationships between the body parts are represented by conditional probability distributions. Pose estimation problem is thus formulated as a probabilistic inference over a graphical model, where the random variables correspond to the individual limb parameters (position and orientation) and Belief Propagation messages ensure coherence between limbs. Secondly, we propose a framework allowing emblematic gestures detection and recognition. The most challenging issue in gesture recognition is to find good features with a discriminant power (to distinguish between different gestures) and a good robustness to intrinsic gestures variability (the context in which gestures are expressed, the morphology of the person, the point of view, etc. ). In this work, we propose a new arm's kinematics normalization scheme reflecting both the muscular activity and arm's appearance when a gesture is performed. The obtained signals are first segmented and then analysed by two machine learning techniques: Hidden Markov Models and Support Vector Machines. The two methods are compared in a 5 classes emblematic gestures recognition task. Both systems show good performances with a minimalistic training database regardless to performer's anthropometry, gender, age or pose with regard to the sensing system. The work presented here has been done within the framework of a PhD thesis in joint supervision between the “Pierre et Marie Curie” University (ISIR laboratory, Paris) and the University of Genova (IIT--Tera department) and was labelled by the French-Italian University

This thesis describes a bottom-up approach to estimating human pose over time based on monocular views with no restriction on human activities,Three approaches are proposed to address the weaknesses of existing approaches, including building a specific appearance model using clustering,utilising both the generic and specific appearance models in the estimation, and building an uncontaminated appearance model by removing backgroundpixels from the training samples. Experimental results show that the proposed system outperforms existing system significantly.

This thesis focuses on the development of a real-time and cost effective marker-less computer vision method for significant body point or part detection (i.e., the head, arm, shoulder, knee, and feet), labelling and tracking, and its application to activity recognition. This work comprises of three parts: significantbody point detection and labelling, significant body point tracking, and activity recognition. Implicit body models are proposed based on human anthropometry, kinesiology, and human vision inspired criteria to detect and label significant body points. The key idea of the proposed method is to fit the knowledge from the implicit body models rather than fitting the predefined models in order to detect and label significant body points. The advantages of this method are that it does not require manual annotation, an explicit fitting procedure, and a training (learning) phase, and it is applicable to humans with different anthropometric proportions. The experimental results show that the proposed method robustly detects and labels significant body points in various activities of two different (low and high) resolution data sets. Furthermore, a Particle Filter with memory and feedback is proposed that combines temporal information of the previous observation and estimation with feedback to track significant body points in occlusion. In addition, in order to overcome the problem presented by the most occluded body part, i.e., the arm, a Motion Flow method is proposed. This method considers the human arm as a pendulum attached to the shoulder joint and defines conjectures to track the arm since it is the most occluded body part. The former method is invoked as default and the latter is used as per a user's choice. The experimental results show that the two proposed methods, i.e., Particle Filter and Motion Flow methods, robustly track significant body points in various activities of the above-mentioned two data sets and also enhance the performance of significant body point detection. A hierarchical relaxed partitioning system is then proposed that employs features extracted from the significant body points for activity recognition when multiple overlaps exist in the feature space. The working principle of the proposed method is based on the relaxed hierarchy (postpone uncertain decisions) and hierarchical strategy (group similar or confusing classes) while partitioning each class at different levels of the hierarchy. The advantages of the proposed method lie in its real-time speed, ease of implementation and extension, and non-intensive training. The experimental results show that it acquires valuable features and outperforms relevant state-of-the-art methods while comparable to other methods, i.e., the holistic and local feature approaches. In this context, the contribution of this thesis is three-fold: Pioneering a method for automated human body part detection and labelling. Developing methods for tracking human body parts in occlusion. Designing a method for robust and efficient human action recognition.

Due to the large influx of computers in our daily lives, human-computer interaction has become crucially important. For a long time, focusing on what users need has been critical for designing interaction methods. However, new perspective tends to extend this attitude to encompass how human desires, interests, and ambitions can be met and supported. This implies that the way we interact with computers should be revisited. Centralizing human values rather than user needs is of the utmost importance for providing new interaction techniques. These values drive our decisions and actions, and are essential to what makes us human. This motivated us to introduce new interaction methods that will support human values, particularly human well-being. The aim of this thesis is to design new interaction methods that will empower human to have a healthy, intuitive, and pleasurable interaction with tomorrow’s digital world. In order to achieve this aim, this research is concerned with developing theories and techniques for exploring interaction methods beyond keyboard and mouse, utilizing human body. Therefore, this thesis addresses a very fundamental problem, human motion analysis. Technical contributions of this thesis introduce computer vision-based, marker-less systems to estimate and analyze body motion. The main focus of this research work is on head and hand motion analysis due to the fact that they are the most frequently used body parts for interacting with computers. This thesis gives an insight into the technical challenges and provides new perspectives and robust techniques for solving the problem.

The purpose of this dissertation is to present our agent-based human tracking framework, and to evaluate the results of our work in light of the previous research in the same field. Our agent-based approach departs from a process-centric model where the agents are bound to specific processes, and introduces a novel model by which agents are bound to the objects or sub-objects being recognized or tracked. The hierarchical agent-based model allows the system to handle a variety of cases, such as single people or multiple people in front of single or stereo cameras. We employ the job-market model for agents' communication. In this dissertation, we will present several experiments in detail, which demonstrate the effectiveness of the agent-based tracking system. Per our research, the agents are designed to be autonomous, self-aware entities that are capable of communicating with other agents to perform tracking within agent coalitions. Each agent with high-level abstracted knowledge seeks evidence for its existence from the low-level features (e.g. motion vector fields, color blobs) and its peers (other agents representing body-parts with which it is compatible). The power of the agent-based approach is its flexibility by which the domain information may be encoded within each agent to produce an overall tracking solution.
Ph. D.

ITC/USA 2008 Conference Proceedings / The Forty-Fourth Annual International Telemetering Conference and Technical Exhibition / October 27-30, 2008 / Town and Country Resort & Convention Center, San Diego, California
This paper describes the design and construction of a low-cost wireless sensor network (WSN) intended to track a human body walking upright through its physical topology. The network consists of arrays of pyroelectric infrared (PIR) sensors that can detect a moving body up to five meters away within a semicircular field of view. Data is gathered from these arrays and transmitted to a central processor that triangulates the body's position. Important characteristics of both the PIR sensors and the network's asynchronous nature are elaborated upon to illustrate how they affect the interpretation of the data.

This thesis addresses the challenging problem of three-dimensional (3-D) reconstruction of a fast-moving human, using a single moving camera which captures both depth and colour information (RGB-D). Our objective is to find solutions to the challenges arising from the high camera motion and articulated human motions. We have developed an effective system which uses the camera pose, skeleton detection, and multi-scale information, to produce a robust reconstruction framework for 3-D modelling of fast-moving humans. The outcome of the research is useful for several applications of human performance capture systems in sports, the arts, and animation industries.

Approved for public release; distribution unlimited
This thesis done in cooperation with the MOVES Institute
Mathematicians have used quaternions for about a hundred years. Today they are an important part of computer graphics and simulation systems. This thesis takes an analytical approach to quaternions by using them in the construction of a virtual human for sourceless Magnetic Accelerometer Rate Sensor (MARG) body tracking system. Virtual citizens will be a reflection of our personalities in cyberspace. Prophecies say they may take control in the virtual world and govern themselves too. One of the objectives of this thesis is to design a seamless and realistic humanoid from laser scan data clouds. This humanoid will be compatible with motion capture systems and networked virtual environments. Second objective of this thesis is to search for the answers related to the optimal real-time representation of an articulated virtual human, maintaining a high level of visual fidelity within networked cyberspace. While visual detail and fidelity have been and will continue to be a major ongoing interest within the computer graphics community, the idea of sourceless body tracking is still in its early stages. MARG body tracking is one of the successful approaches to body tracking systems. This thesis proposes a networked quaternion based real-time virtual human interface for the MARG body tracking system.

Tato diplomová práce byla vytvořena během studijního pobytu na Uviversity of Estern Finland, Joensuu, Finsko. Tato diplomová práce se zabývá využitím technologie sledování pohledu neboli také sledování pohybu očí (Eye-Tracking) pro interakci člověk-počítač (Human-Computer Interaction (HCI)). Navržený a realizovaný systém mapuje pozici bodu pohledu/zájmu (the point of gaze), která odpovídá souřadnicím v souřadnicovém systému kamery scény do souřadnicového systému displeje. Zároveň tento systém kompenzuje pohyby uživatele a tím odstraňuje jeden z hlavních problémů využití sledování pohledu v HCI. Toho je dosaženo díky stanovení transformace mezi projektivním prostorem scény a projektivním prostorem displeje. Za použití význačných bodů (interesting points), které jsou nalezeny a popsány pomocí metody SURF, vyhledání a spárování korespondujících bodů a vypočítání homografie. Systém byl testován s využitím testovacích bodů, které byly rozložené po celé ploše displeje.

Screening individuals for concealed knowledge has traditionally been the purview of professional interrogators investigating a crime. But the ability to detect when a person is hiding important information would be of high value to many other fields and functions. This dissertation proposes design principles for and reports on an implementation and empirical evaluation of a non-invasive, automated system for human screening. The screening system design (termed an automated screening kiosk or ASK) is patterned after a standard interviewing method called the Concealed Information Test (CIT), which is built on theories explaining psychophysiological and behavioral effects of human orienting and defensive responses. As part of testing the ASK proof of concept, I propose and empirically examine alternative indicators of concealed knowledge in a CIT. Specifically, I propose kinesic rigidity as a viable cue, propose and instantiate an automated method for capturing rigidity, and test its viability using a traditional CIT experiment. I also examine oculomotor behavior using a mock security screening experiment using an ASK system design. Participants in this second experiment packed a fake improvised explosive device (IED) in a bag and were screened by an ASK system. Results indicate that the ASK design, if implemented within a highly controlled framework such as the CIT, has potential to overcome barriers to more widespread application of concealed knowledge testing in government and business settings.

The objective of this thesis is to present an algorithm for processing infrared images and accomplishing automatic detection and path tracking of moving subjects with fever. The detection is based on two main features: the distinction between the geometry of a human face and other objects in the field of view of the camera and the temperature of the radiating object. These features are used for tracking the identified person with fever. The position of camera with respect to direction of motion the walkers appeared to be critical in this process. Infrared thermography is a remote sensing technique used to measure temperatures based on emitted infrared radiation. This application may be used for fever screening in major public places such as airports and hospitals. For this study, we first look at human body and objects in a line of view with different temperatures that would be higher than the normal human body temperature (37.8C at morning and 38.3C at evening). As a part of the experimental study, two humans with different body temperatures walking a path were subjected to automatic fever detection applied for tracking the detected human with fever. The algorithm consists of image processing to threshold objects based on the temperature and template matching used for fever detection in a dynamic human environment.

Gait analysis is the systematic study of human walking. Clinical gait analysis is the process by which quantitative information is collected for the assessment and decision-making of any gait disorder. Although observational gait analysis is the therapistâ s primary clinical tool for describing the quality of a patientâ s walking pattern, it can be very unreliable. Modern gait analysis is facilitated through the use of specialized equipment. Currently, accurate gait analysis requires dedicated laboratories with complex settings and highly skilled operators. Wearable locomotion tracking systems are available, but they are not sufficiently accurate for clinical gait analysis. At the same time, wireless healthcare is evolving. Particularly, ultra wideband (UWB) is a promising technology that has the potential for accurate ranging and positioning in dense multi-path environments. Moreover, impulse-radio UWB (IR-UWB) is suitable for low-power and low-cost implementation, which makes it an attractive candidate for wearable, low-cost, and battery-powered health monitoring systems. The goal of this research is to propose and investigate a full-body wireless wearable human locomotion tracking system using UWB radios. Ultimately, the proposed system should be capable of distinguishing between normal and abnormal gait, making it suitable for accurate clinical gait analysis.
Ph. D.

In this dissertation, we address the problem of understanding human activities in videos by developing a two-pronged approach: coarse level modeling of scene activities and fine level modeling of individual activities. At the coarse level, where the resolution of the video is low, we rely on person tracks. At the fine level, richer features are available to identify different parts of the human body, therefore we rely on the body joint tracks. There are three main goals of this dissertation: (1) identify unusual activities at the coarse level, (2) recognize different activities at the fine level, and (3) predict the behavior for synthesizing and tracking activities at the fine level. The first goal is addressed by modeling activities at the coarse level through two novel and complementing approaches. The first approach learns the behavior of individuals by capturing the patterns of motion and size of objects in a compact model. Probability density function (pdf) at each pixel is modeled as a multivariate Gaussian Mixture Model (GMM), which is learnt using unsupervised expectation maximization (EM). In contrast, the second approach learns the interaction of object pairs concurrently present in the scene. This can be useful in detecting more complex activities than those modeled by the first approach. We use a 14-dimensional Kernel Density Estimation (KDE) that captures motion and size of concurrently tracked objects. The proposed models have been successfully used to automatically detect activities like unusual person drop-off and pickup, jaywalking, etc. The second and third goals of modeling human activities at the fine level are addressed by employing concepts from theory of chaos and non-linear dynamical systems. We show that the proposed model is useful for recognition and prediction of the underlying dynamics of human activities. We treat the trajectories of human body joints as the observed time series generated from an underlying dynamical system. The observed data is used to reconstruct a phase (or state) space of appropriate dimension by employing the delay-embedding technique. This transformation is performed without assuming an exact model of the underlying dynamics and provides a characteristic representation that will prove to be vital for recognition and prediction tasks. For recognition, properties of phase space are captured in terms of dynamical and metric invariants, which include the Lyapunov exponent, correlation integral, and correlation dimension. A composite feature vector containing these invariants represents the action and will be used for classification. For prediction, kernel regression is used in the phase space to compute predictions with a specified initial condition. This approach has the advantage of modeling dynamics without making any assumptions about the exact form (polynomial, radial basis, etc.) of the mapping function. We demonstrate the utility of these predictions for human activity synthesis and tracking.
Ph.D.
School of Electrical Engineering and Computer Science
Engineering and Computer Science
Computer Science PhD

La reconstrucció del moviment huma mitjançant l'analisi visual és una area de recerca de la visió per computador plena de reptes amb moltes aplicacions potencials. Els enfocs de seguiment basat en models, i en particular els fltres de partícules, formulen el problema com una tasca d'inferencia Bayesiana l'objectiu de la qual és estimar seqüencialment la distribució sobre els parametres d'un model del cos huma al llarg del temps. Aquests enfocs depenen en gran mesuta d'emprar bons models dinamics i d'observació per tal de predir i actualitzar les confguracions del cos huma en base a mesures extretes de les dades d'imatge. No obstant, resulta molt difícil dissenyar models d'observació, i en especial pel cas de seguiment a partir d'una sola vista, que siguin capaços d'extreure informació útil de les seqüencies d'imatges de manera robusta. Per tant, per tal de superar aquestes limitacions és necessari emprar un fort coneixement a priori sobre el moviment huma i guiar així l'exploració de l'espai d'estats.
El treball presentat en aquesta Tesis esta enfocat a recuperar els parametres de moviment 3D d'un model del cos huma a partir de mesures incompletes i sorolloses d'una seqüencia d'imatges monocular. Aquestes mesures consisteixen en les posicions 2D d'un conjunt redult d'articulacions en el pla d'imatge. Amb aquesta fnalitat, proposem un nou model de moviment huma específc per cada acció, que és entrenat a partir de bases de dades de captures de moviment que contenen varies execucions d'una acció en particular, i que és utilitzat com a coneixement a priori en un esquema de fltratge de partícules.
Les postures del cos es representen emprant un model articulat simple i compacte que fa ús dels cosinus directors per tal de representar la direcció de les parts del cos en l'espai Cartesia 3D. Llavors, donada una acció, s'aplica l'Analisis de Components Principals (PCA) sobre les dades d'entrenament per tal d'aplicar reducció de dimensionalitat sobre les dades d'entrada altament correlacionades. Previament al pas d'entrenament del model d'acció, les seqüencies de moviment d'entrada són sincronitzades mitjançant un nou algoritme d'adaptació dens basat en Programació Dinamica. L'algoritme sincronitza totes les seqüencies de moviment d'una mateixa classe d'acció i és capa¡ de trobar una solució óptima en temps real.
Aleshores, s'apren un model d'acció probabilístic a partir dels exemples de movi¬ment sincronitzats que captura la variabilitat i l'evolució temporal del moviment del cos sencer durant una acció concreta. En particular, per cada acció, els parametres apresos són: una varietat representativa de l'acció que consisteix en l'execució mitjana de la mateixa, la desviació estandard de l'execució mitjana, els vectors de direcció mitjans de cada subseqüencia de moviment d'una llargada donada i l'error esperat en un instant de temps donat.
A continuació, s'utilitza el model específc per cada acció com a coneixement a priori sobre moviment huma que millora l'efciencia i robustesa de tot l'enfoc de seguiment basat en fltratge de partícules. En primer lloc, el model dinamic guia les partícules segons situacions similars apreses previament. A continuació, es restringeix l'espai d'estats per tal que tan sols les postures humanes més factibles siguin acceptades com a solucions valides a cada instant de temps. En conseqüencia, l'espai d'estats és explorat de manera més efcient ja que el conjunt de partícules cobreix les postures del cos més probables.
Finalment, es duen a terme experiments emprant seqüencies de test de varies bases de dades. Els resultats assenyalen que el nostre esquema de seguiment és capa d'estimar la confguració 3D aproximada d'un model de cos sencer, a partir tan sols de les posicions 2D d'un conjunt redult d'articulacions. També s'inclouen proves separades sobre el metode de sincronització de seqüencies i de la tecnica de comparació probabilística de les subseqüencies de moviment.
Recovering human motion by visual analysis is a challenging computer vision research area with a lot of potential applications. Model based tracking approaches, and in particular particle flters, formulate the problem as a Bayesian inference task whose aim is to sequentially estimate the distribution of the parameters of a human body model over time. These approaches strongly rely on good dynamical and observation models to predict and update confgurations of the human body according to mea surements from the image data. However, it is very difcult to design observation models which extract useful and reliable information from image sequences robustly. This results specially challenging in monocular tracking given that only one viewpoint from the scene is available. Therefore, to overcome these limitations strong motion priors are needed to guide the exploration of the state space.
The work presented in this Thesis is aimed to retrieve the 3D motion parameters of a human body model from incomplete and noisy measurements of a monocular image sequence. These measurements consist of the 2D positions of a reduced set of joints in the image plane. Towards this end, we present a novel action specifc model of human motion which is trained from several databases of real motion captured performances of an action, and is used as a priori knowledge within a particle fltering scheme.
Body postures are represented by means of a simple and compact stick fgure model which uses direction cosines to represent the direction of body limbs in the 3D Cartesian space. Then, for a given action, Principal Component Analysis is applied to the training data to perform dimensionality reduction over the highly correlated input data. Before the learning stage of the action model, the input motion performances are synchronized by means of a novel dense matching algorithm based on Dynamic Programming. The algorithm synchronizes all the motion sequences of the same action class, fnding an optimal solution in real time.
Then, a probabilistic action model is learnt, based on the synchronized motion examples, which captures the variability and temporal evolution of full body motion within a specifc action. In particular, for each action, the parameters learnt are: a representative manifold for the action consisting of its mean performance, the stan dard deviation from the mean performance, the mean observed direction vectors from each motion subsequence of a given length and the expected error at a given time instant.
Subsequently, the action specifc model is used as a priori knowledge on human motion which improves the efciency and robustness of the overall particle fltering tracking framework. First, the dynamic model guides the particles according to similar situations previously learnt. Then, the state space is constrained so only feasible human postures are accepted as valid solutions at each time step. As a result, the state space is explored more efciently as the particle set covers the most probable body postures.
Finally, experiments are carried out using test sequences from several motion databases. Results point out that our tracker scheme is able to estimate the rough 3D confguration of a full body model providing only the 2D positions of a reduced set of joints. Separate tests on the sequence synchronization method and the subsequence probabilistic matching technique are also provided.
Keywords: Human Motion Modeling; Particle fltering; Monocular Full Body 3D Tracking.
Topics: Image Processing; Computer Vision; Scene Understanding; Machine Intelligence; Machine Vision Applications; Video-Sequence Evaluation

The range of missions performed by Unmanned Aircraft Systems (UAS) has been steadily growing in the past decades thanks to continued development in several disciplines. The goal of increasing the autonomy of UAS's is widening the range of tasks which can be carried out without, or with minimal, external help. This thesis presents methods for increasing specific aspects of autonomy of UAS's operating both in outdoor and indoor environments where cameras are used as the primary sensors. First, a method for fusing color and thermal images for object detection, geolocation and tracking for UAS's operating primarily outdoors is presented. Specifically, a method for building saliency maps where human body locations are marked as points of interest is described. Such maps can be used in emergency situations to increase the situational awareness of first responders or a robotic system itself. Additionally, the same method is applied to the problem of vehicle tracking. A generated stream of geographical locations of tracked vehicles increases situational awareness by allowing for qualitative reasoning about, for example, vehicles overtaking, entering or leaving crossings. Second, two approaches to the UAS indoor localization problem in the absence of GPS-based positioning are presented. Both use cameras as the main sensors and enable autonomous indoor ight and navigation. The first approach takes advantage of cooperation with a ground robot to provide a UAS with its localization information. The second approach uses marker-based visual pose estimation where all computations are done onboard a small-scale aircraft which additionally increases its autonomy by not relying on external computational power.

Human body is a complex system organized at different levels such as cells, tissues and organs, which contributes to 11 important organ systems. The functional efficiency of this complex system is evaluated as health. Traditional healthcare is unable to accommodate everyone's need due to the ever-increasing population and medical costs. With advancements in technology and medical research, traditional healthcare applications are shaping into smart healthcare solutions. Smart healthcare helps in continuously monitoring our body parameters, which helps in keeping people health-aware. It provides the ability for remote assistance, which helps in utilizing the available resources to maximum potential. The backbone of smart healthcare solutions is Internet of Things (IoT) which increases the computing capacity of the real-world components by using cloud-based solutions. The basic elements of these IoT based smart healthcare solutions are called "things." Things are simple sensors or actuators, which have the capacity to wirelessly connect with each other and to the internet. The research for this dissertation aims in developing architectures for these things, focusing on IoT-based smart healthcare solutions. The core for this dissertation is to contribute to the research in smart healthcare by identifying applications which can be monitored remotely. For this, application-specific thing architectures were proposed based on monitoring a specific body parameter; monitoring physical health for family and friends; and optimizing the power budget of IoT body sensor network using human body communications. The experimental results show promising scope towards improving the quality of life, through needle-less and cost-effective smart healthcare solutions.

碩士
國立成功大學
資訊工程學系碩博士班
98
An auto-tuning template matching for human body tracking is proposed in this thesis. Auto-tuning is incorporated in template matching for our tracking system to evaluate scale change and some occlusion problems of human body. We use the color information to evaluate the most similar region, and to compute the variance between the current template and the reference template based on the concept of median absolute deviation (MAD). According to the variance, the scale of template will be automatically adjusted well. And, the variance is also a standard to evaluate the presence of occlusion. The experimental results show that the proposed tracking system is simple to be implemented and the performance is effective in estimating scale changes and the presence of occlusion.

碩士
淡江大學
資訊工程學系碩士班
101
In recent years, using of Microsoft''s Kinect depth sensor to calculate the human skeleton is called depth skeleton detection. So the way of somatosensory detection can more diverse. Related researches are constantly raised and follow-up, and the action of the detection and analysis is also discussed. There are some motions cannot be detected by the skeleton from depth image, for example, foot-cross and lying will cause judgment failed. In order to improve this misjudgment, this thesis uses the centroid of the color regions to achieve the motion detection. Ensuring the tracking point is the skeleton of the body to achieve the color tracking and solving the problem of misjudgment and tracking. Let the skeleton be tracked correctly.

In recent years with the development of Virtual Reality and gaming industry, a number of Virtual Reality and Motion tracking devices have been offered on the market for an affordable price. Besides the applications in gaming, Virtual Reality potential in the medical rehabilitation was recognized as well. It offers a new approach to treatment in Stroke and Spinal Cord (SCI) Injury rehabilitation. The aim of this work is a research of the application of VR games in the rehabilitation; Identification of how they can increase the motivation of Spinal Cord Injury and Stroke patients for performing exercises relevant for their recovery. This work was performed in collaboration with the Rovisco Pais Rehabilitation Center. Based on the case study of the rehabilitation center and consultancies with the therapists, a set of mini-games was produced. In the first produced mini-game, the aim was training of the upper limb movements for Stroke patients. The developed game yet, hasn’t been tested because of the lack of the occupational therapists. Therefore, the work continued in the direction of producing the mini-game for gait restoring. However, since the evaluation of the Kinect v2 sensor for motion tracking proved that it doesn’t have enough precision the next game developed was a trunk balance training game. The target audience of the game was SCI patients. The produced game was tested with 9 students and based on the results it was further improved and tested with 6 SCI patients. The testing’s results suggested that these types of games can be helpful in the recovery process of SCI patients and can be motivate the patients for the recovery.
Nos últimos anos tem surgido no mercado um grande número de dispositivos de interacção, display e tracking adequados a aplicações de Realidade Virtual, a preços bastante acessíveis, que têm sido usados pela Indústria de Jogos; no entanto, a Realidade Virtual tem também grande potencial na área da Medicina de Reabilitação, podendo oferecer abordagens inovadoras no tratamento de pacientes que recuperam Acidentes Vasculares Cerebrais (AVCs) ou de lesões medulares. O principal objectivo deste trabalho consistiu no estudo da possibilidade de usar aplicações de Realidade Virtual para aumentar a motivação daqueles pacientes na realização continuada de exercícios necessários para a sua recuperação. Este trabalho foi realizado em colaboração com o Centro de Medicina de Reabilitação da Região Centro – Rovisco Pais. Estudaram-se os desafios que os seus médicos, terapeutas e pacientes enfrentam e desenvolveu-se um mini-jogo e adaptou-se um outro para ajudar na recuperação do equilíbrio daqueles pacientes que foi testado primeiro com participantes saudáveis e depois com pacientes. Foi ainda estudada a possibilidade de utilização do sensor Kinect v2 para análise de marcha.
Mestrado em Engenharia de Computadores e Telemática

碩士
國立中興大學
電機工程學系所
107
This thesis proposes a new algorithm to track the two-dimensional shoulders and elbows of a human body using a monocular camera. Based on the located significant points of a human body, this thesis also online builds the skeleton of a human body. In the proposed method, a background is registered by frame difference, and then the foreground (human body) is segmented from the background. The contour of the human body are found by the chain code. The contour and different features proposed in previous studies are used to locate the head, hands, and feet. To locate the elbows, body skeleton is found using the four-point distance transform. The Arcelli-Baja algorithm is then applied to fix the intermittent skeleton. The elbows is localized on the curve connecting the palm and the shoulder based on the skeleton and body contour. In the elbow localization method, different rules are proposed depending different postures of the hand palm. The posture is divided into different cases depending the occupation condition of the palm with body, the position of the palm on the left or right side, and the relative positions of the shoulders and palms. To minimize the tracking error caused by some incorrectly localized elbows and smoothen the trajectory, the particle filter is applied to the localized elbows. Experiments in three videos with comparisons with different elbow tracking methods are performed to verify the effectiveness and accuracy of the proposed elbow tracking method.

碩士
國立成功大學
電腦與通信工程研究所
95
Human body parts tracking plays an important role in computer vision domain. This paper combines local shape context matching with particle filters and using a global constrain of multi-directional kinematic models to track human body parts. Given an input video, the moving object (MO) and its foreground silhouette (FS) are obtained separately by using background subtractions. During the initialization, three main body parts are assigned for both MO and FS first. Then, shape contexts are used to represent for each body part of MO and FS. While applying particle filtering, we found that the position of maximum posterior probability may not be the fittest body parts. By leading in the idea of human kinematics model, the tracked would be more accuracy. To increase the robustness, the skeleton approximating would be applied to track the position of knee and ankle furthermore. Our quantitative analysis of experimental results showed a good performance by using our proposed method.

Hoje em dia, cada vez mais experiências culturais são melhoradas tendo por base aplicações móveis, incluindo aqueles que usam Realidade Aumentada (RA). Estas aplicações têm crescido em número de utilizadores, em muito suportadas no aumento do poder de cálculo dos processadores mais recentes, na popularidade dos dispositivos móveis (com câmaras de alta definição e sistemas de posicionamento global – GPS), e na massificação da disponibilidade de conexões de internet. Tendo este contexto em mente, o projeto Mobile Five Senses Augmented Reality System for Museums (M5SAR) visa desenvolver um sistema de RA para ser um guia em eventos culturais, históricos e em museus, complementando ou substituindo a orientação tradicional dada pelos guias ou mapas. O trabalho descrito na presente tese faz parte do projeto M5SAR. O sistema completo consiste numa aplicação para dispositivos móveis e num dispositivo físico, a acoplar ao dispositivo móvel, que em conjunto visam explorar os 5 sentidos humanos: visão, audição, tato, olfacto e paladar. O projeto M5SAR tem como objetivos principais (a) detectar peças do museu (por exemplo, pinturas e estátuas (Pereira et al., 2017)), (b) detectar paredes / ambientes do museu (Veiga et al., 2017) e (c) detectar formas humanas para sobrepor o conteúdo de Realidade Aumentada (?). Esta tese apresenta uma abordagem relativamente ao último objectivo, combinando informações de articulações do corpo humano com métodos de sobreposição de roupas. Os atuais sistemas relacionados com a sobreposição de roupas, que permitem ao utilizador mover-se livremente, são baseados em sensores tridimensionais (3D), e.g., Sensor Kinect (Erra et al., 2018), sendo estes não portáteis. A contribuição desta tese é apresentar uma solução portátil baseado na câmara (RGB) do telemóvel que permite ao utilizador movimentar-se livremente, fazendo ao mesmo tempo a sobreposição de roupa (para o corpo completo). Nos últimos anos, a capacidade de Redes Neurais Convolucionais (CNN) foi comprovado numa grande variedade de tarefas de visão computacional, tais como classificação e detecção de objetos e no reconhecimento de faces e texto (Amos et al., 2016; Ren et al., 2015a). Uma das áreas de uso das CNN é a estimativa de posição (pose) humana em ambientes reais (Insafutdinov et al., 2017; Pishchulin et al., 2016). Recentemente, duas populares CNN frameworks para detecção e segmentação de formas humanas apresentam destaque, o OpenPose (Cao et al., 2017;Wei et al., 2016) e o Mask R-CNN (He et al., 2017). No entanto, testes experimentais mostraram que as implementações originais não são adequadas para dispositivos móveis. Apesar disso, estas frameworks são a base para as implementações mais recentes, que possibilitam o uso em dispositivos móveis. Uma abordagem que alcança a estimativa e a segmentação de pose de corpo inteiro é o Mask R-CNN2Go (Jindal, 2018), baseado na estrutura original do Mask R-CNN. A principal razão para o tempo de processamento ser reduzido foi a otimização do número de camadas de convolução e a largura de cada camada. Outra abordagem para obter a estimativa de pose humana em dispositivos móveis foi a modificação da arquitetura original do OpenPose para mobile (Kim, 2018; Solano, 2018) e sua combinação com MobileNets (Howard et al., 2017). MobileNets, como o nome sugere, é projetado para aplicativos móveis, fazendo uso de camadas de convoluções separáveis em profundidade. Essa modificação reduz o tempo de processamento, mas também reduz a precisão na estimativa da pose, quando comparado à arquitetura original. É importante ressaltar que apesar de a detecção de pessoas com a sobreposição de roupas ser um tema atual, já existem aplicações disponíveis no mercado, como o Pozus (GENTLEMINDS, 2018). O Pozus é disponibilizado numa versão beta que é executado no sistema operativo iOS, usa a câmera do telemóvel como entrada para a estimação da pose humana aplicando segmentos de texturas sobre o corpo humano. No entanto, Pozus não faz ajuste de texturas (roupas) à forma da pessoa. Na presente tese, o modelo OpenPose foi usado para determinar as articulações do corpo e diferentes abordagens foram usadas para sobreposição de roupas, enquanto uma pessoa se move em ambientes reais. A primeira abordagem utiliza o algoritmo GrabCut (Rother et al., 2004) para segmentação de pessoas, permitindo o ajuste de segmentos de roupas. Uma segunda abordagem usa uma ferramenta bidimensional (2D) de Animação do Esqueleto para permitir deformações em texturas 2D de acordo com as poses estimadas. A terceira abordagem é semelhante à anterior, mas usa modelos 3D, volumes, para obter uma simulação mais realista do processo de sobreposição de roupas. Os resultados e a prova de conceito são mostrados. Os resultados são coerentes com uma prova de conceito. Os testes revelaram que como trabalho futuro as otimizações para melhorar a precisão do modelo de estimação da pose e o tempo de execução ainda são necessárias para dispositivos móveis. O método final utilizado para sobrepor roupas no corpo demonstrou resultados positivos, pois possibilitaram uma simulação mais realística do processo de sobreposição de roupas.
When it comes to visitors at museums and heritage places, objects speak for themselves. Nevertheless, it is important to give visitors the best experience possible, this will lead to an increase in the visits number and enhance the perception and value of the organization. With the aim of enhancing a traditional museum visit, a mobile Augmented Reality (AR) framework is being developed as part of the Mobile Five Senses Augmented Reality (M5SAR) project. This thesis presents an initial approach to human shape detection and AR content superimposition in a mobile environment, achieved by combining information of human body joints with clothes overlapping methods. The present existing systems related to clothes overlapping, that allow the user to move freely, are based mainly in three-dimensional (3D) sensors (e.g., Kinect sensor (Erra et al., 2018)), making them far from being portable. The contribution of this thesis is to present a portable system that allows the user to move freely and does full body clothes overlapping. The OpenPose model (Kim, 2018; Solano, 2018) was used to compute the body joints and different approaches were used for clothes overlapping, while a person is moving in real environments. The first approach uses GrabCut algorithm (Rother et al., 2004) for person segmentation, allowing to fit clothes segments. A second approach uses a bi-dimensional (2D) skeletal animation tool to allow deformations on 2D textures according to the estimated poses. The third approach is similar to the previous, but uses 3D clothes models (volumes) to achieve a more realistic simulation of the process of clothes superimposition. Results and proof-of-concept are shown.

碩士
淡江大學
資訊工程學系碩士班
101
The common solution of human-computer interaction is used depth sensor to establish human skeleton, but the skeleton information will be incorrect in some action such as sitting posture, lying posture, and legs crossed. The first two posture that human body is too close with background object, the sensor cannot recognize human and background. The last one, leg crossed, because cannot recognize the relationship of crossing leg. When people in one of these three posture, the skeleton tracking will be failed. In this thesis, we focus on fixing the bug on legs crossed posture. To solve this problem, we proposes a depth image of the skeleton tracking, when a cross-leg, then the color images be used as an aid to re-establish the skeleton, so as to maintain the correct tracking results.

博士
國立交通大學
資訊科學與工程研究所
100
The advance of sensing technology and wireless communication has boosted body-area inertial sensor networks (BISNs), in which wireless wearable inertial sensor nodes are deployed on a human body to monitor its motion. Applications include medical care, pervasive video games, and affective computing. We conduct fundamental research into the technologies required to create an efficient wireless communication BISN that maximizes motion tracking accuracy and data collection efficiency. The first work addresses data collection issues in BISNs by data compression. We observe that, when body parts move, although sensor nodes in vicinity may compete strongly with each other, the transmitted data usually exists some levels of redundancy and even strong temporal and spatial correlations. Our scheme is specifically designed for BISNs, where nodes are likely fully connected and overhearing among sensor nodes is possible. We model the data compression problem for BISNs, where overhearing should be efficiently utilized, as a combinatorial optimization problem on overhearing graphs. We show its computational complexity and present efficient algorithms. We also discuss the design of the underlying MAC protocol to support our compression model. An experimental case study in Pilates exercises for patient rehabilitation is reported. The results show that our schemes reduce more than 70% of overall transmitted data compared with existing approaches. Based on the first work, where a node is allowed to overhear at most $\kappa = 1$ node's transmission, in the second work, we consider multi-spatial correlations by extending $\kappa = 1$ to $\kappa > 1$ and constructing a partial-ordering directed acyclic graph (DAG) to represent the compression dependencies among sensor nodes. While a minimum-cost tree for $\kappa = 1$ can be found in polynomial time, we show that finding a minimum-cost DAG is NP-hard even for $\kappa = 2$. We then propose an efficient heuristic and verify its performance by real sensing data. In addition to data collection, in the third work, we are interested in tracking human postures by deploying accelerometers on a human body. One fundamental issue in such scenarios is how to calculate the gravity. This is very challenging especially when the human body parts keep on moving. Assuming multiple accelerometers being deployed on a rigid part of a human body, a recent work proposes a data fusion method to estimate the gravity vector on that rigid part. However, how to find the optimal deployment of sensors that minimizes the estimation error of the gravity vector is not addressed. In this work, we formulate the deployment optimization problem and propose two heuristics, called Metropolis-based method and largest-inter-distance-based (LID-based) method. Simulation and real experimental results show that our schemes are quite effective in finding near-optimal solutions for a variety of rigid body geometries.

M.TECH
The motivation behind this project is to develop software for tracking and recognizing the human activity major application in security, surveillance and vision analysis. The developed software must be capable of tracking the human body and recognizing its activity. The proposed method uses the approach for features extraction from the sequences of images. The method describes about the recognition of human activity with the help of change in energy produced by motion of the connected pixels in an image and then we used the support vector machine as the classifier. The proposed technique takes care of the real time implementation of the technique and in qualitative decision making both and shows better results. This technique is capable of understanding the activity. The statistical confidence is higher as compared to the previous techniques because the activity recognition is based upon the features of not just one organ but also on the dependent organs. This method works in real time and is inherently parallel.

碩士
國立中正大學
資訊工程研究所
105
As technology advances, the wearable personal entertainments and personal healthcare devices are booming. Traditionally, the medical healthcare devices such as electromyography (EMG) and electrocardiography (ECG) use the wireline to transfer the physiological signals, it leads to inconvenient for patients. Therefore, the wireless communication techniques are used to solve this problem. The common type of wireless body area network (WBAN) realization transmits data through the air, such as Bluetooth and ZigBee. However, they have relatively high power consumption and relatively low data rate. Therefore, the body channel communication (BCC) is proposed to solve this problem. The wireless communication transmits the data through the air, but the BCC uses the human body as the communication channel to transmit the data. Moreover, the BCC has relatively low signal attenuation and relatively few interferences in the nearby environment. Furthermore, the BCC can achieve a high data rate. In this thesis, a wideband signaling (WBS) transceiver is proposed. In transmitter part, the data are modulated by a non-return to zero inverted (NRZI) encoder with a bit stuffer. The data are transmitted in the packet format. In the receiver part, the proposed reference-less clock and data recovery circuit (CDR) can reduce the power consumption and circuit complexity. Moreover, the proposed phase error calculation method uses the multi-phase signals to quantize the phase error. The proposed method can enhance the CDR phase tracking ability and compensate for phase error quickly. The proposed automatically phase track gain calibration method can calculate the gain value for the all-digital CDR (ADCDR) controller. In different process, voltage and temperature (PVT) variations, the gain value will be automatically calibrated.

ABSTRACT The motivation behind this project is to develop software for tracking and recognizing the human activity major application in security, surveillance and vision analysis. The developed software must be capable of tracking the human body and recognizing its activity. The proposed method uses the approach for features extraction from the sequences of images. The method describes about the recognition of human activity with the help of change in energy produced by motion of the connected pixels in an image and then we used the support vector machine as the classifier. The proposed technique takes care of the real time implementation of the technique and in qualitative decision making both and shows better results. This technique is capable of understanding the activity. The statistical confidence is higher as compared to the previous techniques because the activity recognition is based upon the features of not just one organ but also on the dependent organs. This method works in real time and is inherently parallel

Due to low birth rates and rising life expectancy, the population of developed countries is aging. Concurrently, physical inactivity is an identified major health risk and significantly more prevalent in older adults, who experience the consequences related to inactivity more frequently. Exergames for the elderly are an affordable option to prevent sedentarism and complement traditional exercise training, which can otherwise suffer from low adherence and personalization. They facilitate moderate-intensity physical activity levels and positively impact fitness, health, balance, postural control, mobility, and motivation. However, due to the lack of knowledge of seniors’ game preferences and technology literacy, there are challenges in designing exergames that match the users’ needs and motivators with game elements. While there is an extensive body of research in this field, there are critical gaps: most of the research is done in laboratory environments, is focused on balance and ignore other motor performance domains, use commercial games which are not designed for older adults, and fail to explore the longitudinal effects of exergames. In this thesis, there are three sequential contributions: 1) Develop a technology to facilitate exergaming in the elderly population by integrating easy-to-use full body interaction with large projection displays. Resulting in software for low-cost virtual reality surround screen projection systems, validated through user studies and compared with the conventional alternatives. 2) Leverage the technology and design customized exergames to promote fitness in older adults by: a) Evaluate the capability to automate fitness assessment using gesture detectors by testing their performance in the field with 22 elderly end-users and compare it to traditional methods administered by an expert. Resulting in a high accuracy system, consistent with the traditional fitness assessment method. b) Study older adults’ interaction preferences with floor projection displays by developing and testing two natural user interfaces with 19 elderly participants. The participants’ preference for a feet controlled interface was identified when usability, perceived workload, and performance indicators were assessed. c) Apply human-centered design methodologies in the gamification of fitness training routines by focusing on insights from inquiries to improve game elements and game iterations based on playtesting sessions to produce exergames. Resulting in a set of four exergames created to train the critical functional fitness areas of older adults. 3) Measure the benefits in older adults’ motor performance, quality of life, and physical activity levels during a longitudinal multidimensional training combining custom-made exergames and traditional exercise in a complementary fashion. Achieved through a 12-week long randomized controlled trial of bi-weekly exercise sessions with 31 elderly participants. Outcome measures on fitness, balance, and health-related quality of life were measured at the start, during, and after the intervention, and physical activity levels were measured at each session. This resulted in exergame players having a significant increase in strength compared to control, and both conditions improving balance and the mental component of health-related quality of life, with improvements in the latter being greater for exergame players. Additionally, during exergames’ sessions, participants spent less energy but maintained the recommended physical activity levels for more extended periods. Our results show that integrating personalized exergames designed for multidimensional fitness training in traditional settings can effectively enhance older adults’ motor performance and mental well-being. This technology is a viable low-cost option to be deployed in the context of elderly fitness programs.
Devido às baixas taxas de natalidade e ao aumento da esperança média de vida a população dos países desenvolvidos está a tornar-se envelhecida. Ao mesmo tempo, a falta de atividade física está identificada como um importante fator de risco para a saúde, com grande prevalência em idosos, que também experienciam as suas consequências com maior frequência. Os exergames para idosos são uma opção económica para a prevenção de sedentarismo e complemento do treino físico tradicional, que pode sofrer de baixa aderência e personalização. Estes jogos promovem níveis moderados de intensidade da atividade física e benefícios ao nível do fitness, saúde, equilíbrio, postura, mobilidade e motivação. No entanto, devido ao desconhecimento sobre as preferências de jogos e literacia tecnológica dos idosos, há desafios no desenho de exergames que se adequem às suas necessidades e motivações. Apesar de haver extensa investigação nesta área, existem lacunas críticas: a maior parte da investigação é feita em ambiente laboratorial, foca-se no equilíbrio e ignora os outros domínios motores, usa jogos comerciais que não foram desenhados para esta população e não explora os efeitos longitudinais dos exergames. Nesta tese, fazemos três contribuições, apresentadas de forma sequencial: 1) Desenvolvimento de tecnologia para a prática de exergaming pela população idosa, integrando interação corporal fácil de usar com projeções de grandes dimensões. Resultando num software para uso em sistemas de realidade virtual através de projeção de baixo custo, validado através de estudos com utilizadores e comparado com alternativas convencionais. 2) Uso da tecnologia e design de exergames feitos à medida para a promoção do fitness em idosos: a) Apreciação da capacidade de automação da avaliação do fitness através do uso de detetores de gestos, testando o seu desempenho no terreno com 22 utilizadores idosos. Resultando num sistema de elevada exatidão, consistente com os métodos tradicionais. b) Estudo das preferências de idosos na interação com projeções no solo, através do desenvolvimento e teste de dois interfaces naturais com 19 participantes idosos. Identificada uma preferência pelo interface controlado pelos pés através da avaliação de usabilidade, carga de trabalho sentida e indicadores de desempenho. c) Aplicação de metodologias de desenho centrado em humanos à gamificação de rotinas de treino físico através do foco na compreensão de inquéritos, baseados em sessões de jogo, aplicados para a melhoria e iteração dos mesmos. Resultando num conjunto de quatro exergames criados para treinar áreas críticas de fitness funcional em idosos. 3) Medição do desempenho motor, qualidade de vida e intensidade da atividade física de idosos durante treino longitudinal multidimensional combinando exergames feitos à medida com exercício tradicional de forma complementar. Alcançada através de um estudo randomizado controlado de 12 semanas, com sessões de exercício físico bissemanais por parte de 31 participantes idosos. Com medições de fitness, equilíbrio e qualidade de vida relacionada com a saúde medidas no início, durante e após a intervenção, e níveis de atividade física medidos em todas as sessões. Resultando num aumento significativo de força nos jogadores de exergames relativamente ao controlo e numa melhoria em ambos os grupos no equilíbrio e na componente mental da qualidade de vida, com uma melhoria maior por parte dos jogadores nesta última. Adicionalmente, registado um menor dispêndio energético, mas uma manutenção mais prolongada da intensidade da atividade física recomendada nas sessões de exergames. Os nossos resultados mostram que a integração de exergames personalizados, desenhados para o treino multidimensional de fitness, em ambientes tradicionais de treino podem efetivamente melhorar os ganhos de desempenho motor e bem-estar mental. Assim, esta tecnologia é uma opção económica viável para ser usada no contexto de programas de fitness para idosos.