Thesis (Ph. D.)--University of Washington, 2008. / Vita. Includes bibliographical references (p. 105-112).

Fundamental movement skills (FMSs) are considered to be one of the essential phases of motor skill development. The proper development of FMSs allows children to participate in more advanced forms of movements and sports. To be able to perform an FMS correctly, children need to learn the right way of performing it. By making use of technology, a system can be developed that can help facilitate the learning of FMSs. The objective of the research was to propose an effective natural user interface (NUI) architecture for detecting FMSs using the Kinect. In order to achieve the stated objective, an investigation into FMSs and the challenges faced when teaching them was presented. An investigation into NUIs was also presented including the merits of the Kinect as the most appropriate device to be used to facilitate the detection of an FMS. An NUI architecture was proposed that uses the Kinect to facilitate the detection of an FMS. A framework was implemented from the design of the architecture. The successful implementation of the framework provides evidence that the design of the proposed architecture is feasible. An instance of the framework incorporating the jump FMS was used as a case study in the development of a prototype that detects the correct and incorrect performance of a jump. The evaluation of the prototype proved the following: - The developed prototype was effective in detecting the correct and incorrect performance of the jump FMS; and - The implemented framework was robust for the incorporation of an FMS. The successful implementation of the prototype shows that an effective NUI architecture using the Kinect can be used to facilitate the detection of FMSs. The proposed architecture provides a structured way of developing a system using the Kinect to facilitate the detection of FMSs. This allows developers to add future FMSs to the system. This dissertation therefore makes the following contributions: - An experimental design to evaluate the effectiveness of a prototype that detects FMSs - A robust framework that incorporates FMSs; and - An effective NUI architecture to facilitate the detection of fundamental movement skills using the Kinect.

Modern smartphones contain multiple sensors and long lasting batteries, making them ideal platforms for mobility monitoring. Mobility monitoring can provide rehabilitation professionals with an objective portrait of a patient’s daily mobility habits outside of a clinical setting. The objective of this thesis was to improve the performance of the human activity recognition within a custom Wearable Mobility Measurement System (WMMS). Performance of a current WMMS was evaluated on able-bodied and stroke participants to identify areas in need of improvement and differences between populations. Signal features for the waist-worn smartphone WMMS were selected using classifier-independent methods to identify features that were useful across populations. The newly selected features and a transition state recognition method were then implemented before evaluating the improved WMMS system’s activity recognition performance. This thesis demonstrated: 1) diverse population data is important for WMMS system design; 2) certain signal features are useful for human activity recognition across diverse populations; 3) the use of carefully selected features and transition state identification can provide accurate human activity recognition results without computationally complex methods.

This thesis presents contributions to the theory and practice of team activity recognition. A particular focus of our work was to improve our ability to collect and label representative samples, thus making the team activity recognition more efficient. A second focus of our work is improving the robustness of the recognition process in the presence of noisy and distorted data. The main contributions of this thesis are as follows: We developed a software tool, the Teamwork Scenario Editor (TSE), for the acquisition, segmentation and labeling of teamwork data. Using the TSE we acquired a corpus of labeled team actions both from synthetic and real world sources. We developed an approach through which representations of idealized team actions can be acquired in form of Hidden Markov Models which are trained using a small set of representative examples segmented and labeled with the TSE. We developed set of team-oriented feature functions, which extract discrete features from the high-dimensional continuous data. The features were chosen such that they mimic the features used by humans when recognizing teamwork actions. We developed a technique to recognize the likely roles played by agents in teams even before the team action was recognized. Through experimental studies we show that the feature functions and role recognition module significantly increase the recognition accuracy, while allowing arbitrary shuffled inputs and noisy data.

No description available.

Electronic textiles provide an effective platform to contain wearable computing elements, especially components geared towards the application of activity recognition. An activity recogni tion system built into a wearable textile substrate can be utilized in a variety of areas including health monitoring, military applications, entertainment, and fashion. Many of the activity recognition and motion capture systems previously developed have several drawbacks and limitations with regard to their respective designs and implementations. Some such systems are often times expensive, not conducive to mass production, and may be difficult to calibrate. An effective system must also be scalable and should be deployable in a variety of environments and contexts. This thesis presents the design and implementation of a self-contained motion sensing wearable electronic textile system with an emphasis toward the application of activity recognition. The system is developed with scalability and deployability in mind, and as such, utilizes a two-tier hierarchical model combined with a network infrastructure and wireless connectivity. An example prototype system, in the form of a jumpsuit garment, is presented and is constructed from relatively inexpensive components and materials. / Master of Science

No description available.

This thesis concerns the application of different artificial neural network architectures on the classification of multivariate accelerometer time series data into activity classes such as sitting, lying down, running, or walking. There is a strong correlation between increased health risks in children and their amount of daily screen time (as reported in questionnaires). The dependency is not clearly understood, as there are no such dependencies reported when the sedentary (idle) time is measured objectively. Consequently, there is an interest from the medical side to be able to perform such objective measurements. To enable large studies the measurement equipment should ideally be low-cost and non-intrusive. The report investigates how well these movement patterns can be distinguished given a certain measurement setup and a certain network structure, and how well the networks generalise to noisier data. Recurrent neural networks are given extra attention among the different networks, since they are considered well suited for data of sequential nature. Close to state-of-the-art results (95% weighted F1-score) are obtained for the tasks with 4 and 5 classes, which is notable since a considerably smaller number of sensors is used than in the previously published results. Another contribution of this thesis is that a new labeled dataset with 12 activity categories is provided, consisting of around 6 hours of recordings, comparable in number of samples to benchmarking datasets. The data collection was made in collaboration with the Department of Public Health at Karolinska Institutet. / Inom ramen för uppsatsen testas hur väl rörelsemönster kan urskiljas ur accelerometerdatamed hjälp av den gren av maskininlärning som kallas djupinlärning; där djupa artificiellaneurala nätverk av noder funktionsapproximerar mappandes från domänen av sensordatatill olika fördefinerade kategorier av aktiviteter så som gång, stående, sittande eller liggande.Det finns ett intresse från den medicinska sidan att kunna mäta fysisk aktivitet objektivt,bland annat eftersom det visats att det finns en korrelation mellan ökade hälsorisker hosbarn och deras mängd daglig skärmtid. Denna typ av mätningar ska helst kunna göras medicke-invasiv utrustning till låg kostnad för att kunna göra större studier.Enklare nätverksarkitekturer samt återimplementeringar av bästa möjliga teknik inomområdet Mänsklig aktivitetsigenkänning (HAR) testas både på ett benchmarkingdataset ochpå egeninhämtad data i samarbete med Institutet för Folkhälsovetenskap på Karolinska Institutetoch resultat redovisas för olika val av möjliga klassificeringar och olika antal dimensionerper mätpunkt. De uppnådda resultaten (95% F1-score) på ett 4- och 5-klass-problem ärjämförbara med de bästa tidigare publicerade resultaten för aktivitetsigenkänning, vilket äranmärkningsvärt då då betydligt färre accelerometrar har använts här än i de åsyftade studierna.Förutom klassificeringsresultaten som redovisas bidrar det här arbetet med ett nyttinhämtat och kategorimärkt dataset; KTH-KI-AA. Det är jämförbart i antal datapunkter medspridda benchmarkingdataset inom HAR-området.

The goal of human action recognition on videos is to determine in an automatic way what is happening in a video. This work focuses on providing an answer to this question: given consecutive frames from a video where a person or persons are doing an action, is an automatic system able to recognize the action that is going on for each person? Seven approaches have been provided, most of them based on an alignment process in order to find a measure of distance or similarity for obtaining the classification. Some are based on fluents that are converted to qualitative sequences of Allen relations to make it possible to measure the distance between the pair of sequences by aligning them. The fluents are generated in various ways: representation based on feature extraction of human pose propositions in just an image or a small sequence of images, changes of time series mainly on the angle of slope, changes of the time series focus on the slope direction, and propositions based on symbolic sequences generated by SAX. Another approach based on alignment corresponds to Dynamic Time Warping on subsets of highly dependent parts of the body. An additional approach explored is based on SAX symbolic sequences and respective pair wise alignment. The last approach is based on discretization of the multivariate time series, but instead of alignment, a spectrum kernel and SVM are used as is employed to classify protein sequences in biology. Finally, a sliding window method is used to recognize the actions along the video. These approaches were tested on three datasets derived from RGB-D cameras (e.g., Microsoft Kinect) as well as ordinary video, and a selection of the approaches was compared to the results of other researchers.

Building smart home environments which automatically or semi-automatically assist and comfort occupants is an important topic in the pervasive computing field, especially with the coming of cheap, easy-to-install sensors. This has given rise to the indispensable need for human activity recognition from ubiquitous sensors whose purpose is to observe and understand what occupants are trying to do from sensory data. The main approach to the problem of human activity recognition is a probabilistic one so as to handle the complication of uncertainty, the overlapping of human behaviours and environmental noise. This thesis develops a probabilistic model as a framework for human activity recognition using multiple multi-modal sensors in complex pervasive environments. The probabilistic model to be developed is adapted and based on the abstract hidden Markov model (AHMM) with one layer to fuse multiple sensors. The concept of factored state representation is employed in the model to parsimoniously represent the state transitions for reducing the number of required parameters. The exact method is used in learning the model’s parameters and performing inference. To be able to incorporate a large number of sensors, several more parsimonious representations including the mixtures of smaller multinomials and sigmoid functions are investigated to model the state transitions, resulting in a reduction of the number of parameters and time required for training. / We examine the approximate variational method to significantly reduce the time required for training the model instead of using the exact method. A system of fixed point equations is derived to iteratively update the free variational parameters. We also present the factored model in the case where all variables are continuous with the use of the conditional Gaussian distribution to model state transitions. The variational method is still employed in this case to speed up the model’s training process. The developed model is implemented and applied in recognizing daily activity in our smart home and the Nokia lab from multiple sensors. The experimental results show that the model is appropriate for fusing multiple sensors in activity recognition with a reasonable recognition performance.