As computing becomes increasingly embedded into the fabric of everyday life, systems that understand people’s context of use are of paramount importance. Regardless of whether the platform is a mobile device, a wearable, or part of the “Internet of Things", context offers an implicit dimension that is vital in reducing interactive viscosity between tasks and increasing the richness of human-computer interactions. Sensors are the primary interface for bringing context awareness, but existing sensing approaches are often costly, obtrusive, and special-purpose. Numerous approaches have been attempted and articulated, though none have reached widespread use to date.
Motivated by this problem, my thesis work focuses on enhancing context-awareness through ubiquitous and unobtrusive sensing, drawing upon machine learning to unlock a wide range of applications. I attack this problem space on two fronts: 1) using wearables to transform the human arm into a context-sensing springboard, and 2) transforming everyday spaces into smart environments with general-purpose sensors. The systems that I built have been deployed across long periods and multiple environments, the results of which show the versatility, accuracy and potential for practical context sensing. By combining novel sensing with machine learning, my work transforms raw signals into intelligent abstractions that can power rich, context-sensitive applications, unleashing the potential of next-generation computing platforms.
Chris Harrison (Chair)
Shwetak Patel (University of Washington)