My research interests are in the application of innovative technology-driven techniques for assessing, monitoring, and intervening in the lives of individuals living with chronic conditions. This includes the development and validation of models of recovery in individuals with chronic disabilities using novel technologies, including wearable sensors and assistive robotics. The central motivating theme of my research is ambient assisted living: the use of assistive technologies in real-world settings outside of clinical or laboratory environments to assist people living with disability. My ongoing areas of research include:
Human-robot Interaction (HRI) for Therapeutic Rehabilitation: Task oriented training is known to be one of the dominant techniques for motor neurorehabilitation after stroke. However, many features of such practice (e.g. content and quantity of feedback, practice schedules, motivation) may impact the outcomes. HRI presents a novel framework wherein the interaction of these variables can be systematically evaluated in order to elicit better outcomes from participants. We have been developing algorithms for a socially assistive robotic (SAR) agent to supervise and guide participants as they perform a therapy-inspired motor rehabilitative task. We combine evidence guided theories of motivation, self-efficacy, and challenge point with user state estimates, as determined by classification of multimodal sensor data. Our interests are in analyzing the relationship between the system controller and the user experience and performance during therapeutic task inspired interactions.
Activity Recognition and Motion Analysis: Functional motor capability can change over various time periods, whether it is in response to chronic event such as stroke, after an individual with Parkinson's disease takes medication, or over the time course of aging. These changes in capability may reflect the progression of recovery, changes in motor control, or changes in movement strategy. We are developing techniques that use simple, non-invasive motion sensors to ascertain functional level and pinpoint characteristics that can be used to quantify quality of motion (such as directness of path, jerkiness, and velocity). Further, we are using spectral analyses of motion data to reveal different types of motion characteristics in a variety of controlled (e.g. laboratory) and ambient (e.g. in the home) environments. Observing the progression of such measures outside of the clinical environment will provide insight into the mechanisms of change in such individuals.
Automation of Functional Motor Assessments: Functional outcome assessments are used to determine the efficacy of clinical interventions. Such assessments continue to suffer from limited reliability. We are developing tools to automate standard outcome measures for functional ability. To do so, we collect clinician rated scores of performance in individuals with chronic disability while also measuring their motion. Subsequently, we use Bayesian techniques to predict the scores based on kinematic and dynamic measures obtained from wearable sensors. This will lead to more reliable administration of assessments and better decision support for interveners.
Wearable Sensor Design: Wearable sensors and body sensor networks allow us to collect meaningful physiological data in ambient environments, and over longitudinal time periods. However, designing devices to be used in such environments requires expertise in mechanical design, human factors, smart materials, and electromechanical modeling techniques. I have been developing sensor hardware and data analysis algorithms to process wearable sensor data. I am interested in combining multiple sensor modalities to determine better estimates of user state that will provide insight into the mechanisms of recovery and change due to chronic conditions.
We would like to acknowledge our research support from: