Empath: A System for the Collection, Storage, Analysis, and Reporting of Objective Behavioral Measures

Abstract

Remote health monitoring deployed in homes could help streamline the efficiency of the medical system by decreasing the costs of senior care and providing preventative care to keep people out of hospitals. Besides for caring for a particular person, these systems will generate tremendous amount of aggregate data that can be used to help diagnose diseases and run longitudinal studies on epidemiology. Most systems are not yet flexible enough to handle adding additional modalities into the system or to leverage the cloud's horizontal scalablity for storage, analysis, and display of this data. is dissertation presents a novel framework that sets itself apart from existing remote health monitoring systems for its scalability and flexibility. In addition, this work advances two key technologies for home healthcare. e first is the creation and evaluation of cheap and noninvasive sleep monitoring systems and novel accelerometor-based systems from RFIDs and motes.Secondly, this thesis demonstrates that the general features from speech in the home can be a useful modality for measuring social interaction and mood and promotes a solution to existing technical problems. Ambient monitoring of speech in the home has not been successful primarily because distortion from a room's acoustics negatively impacts classification results. is work presents a novel matched-condition classifier using cuboid acoustic simulation to achieve accuracy comparable to ideal close-to-microphone conditions. Most health monitoring systems have only been tested in a lab and under very scripted scenarios. is system, however, has been used in three different applications: monitoring sleep behaviors and stress for those who suffer from severe epilepsy in a clinical study, another clinical application that studies the relationship of incontinence with sleep agitation for those suffering from Alzheimer's disease, and an in-home deployment monitoring i Abstract ii important factors of depression. We present the commonalities among these different applications and show how to adapt the system for these purposes. AcknowledgementsTo Valerie-for editing my incoherent writing, volunteering to be instrumented by all sorts of sensors, and keeping me sane these past few months. I love you! To Mom and Dad-for teaching me stay "well balanced" and to always remain intellectually curious.To Jack-who has supported me for many years and has taught me the importance of achieving excellence. Without your help, I would never have become a professor.To Enamul-without your help on all of the projects-no data would have been collected.To Ann, Jewel, Karen, and Windy-for finding patients and putting Empath into real studies.To Ben-for taking a risk with taking in an undergraduate, teaching me to be a good mentor, and encouraging me to go to graduate school. More thanks to