Abstract
In today's context, the term 'Smart Object' is used to refer to any object that incorporates computing and communication in some capacity to enhance the functionality and/or interaction experience for the end user. People interacting with objects have different personalities and preferences, and therefore different requirements from the same objects. Therefore, personalization of a Smart Object's behavior will soon become an important function. Often when people interact with objects that are shared with other people, they need to re-configure the objects to suit their personal requirements. For example, while showering, people prefer different water temperatures, and therefore have to set the hot and cold water mixer to their preferred configuration every time they shower.
In the current state of the art, if the Smart Objects need to adapt themselves to the person using them, they perform an explicit identification process for the user. However as Smart Objects are expected to permeate every aspect of people's daily lives, this approach will not be scalable. In order for objects to perform personalized functions, they must solve what we refer to as the Implicit Object User Identification (IOUI) problem: understanding who is actually using a given object, and being able to validate their identity without expecting the user to explicitly participate in an identification process.
In this dissertation, we explore the use of wearable devices in performing IOUI. There are two main reasons why wearable devices are an attractive technological solution that can assist Smart Objects in solving this problem: a) wearables adoption is growing at a rapid pace, and b) they are embedded with sensors that can monitor the location context and hand motion of the wearer. However, while sensors in wearables are great for making approximate measures of a person's activities, the imprecision of their sensing systems makes them challenging for use in applications such as IOUI, which require high precision and accuracy.
In this work, we explore the following hypothesis: Despite the coarse granularity of its location sensing, and imprecision in sensing the trajectory of hand's motion, data from sensors in wearable devices, when augmented with data from Smart Objects, can be used to identify users interacting with Smart Objects. We explore different levels of information shared by the Wearable device with a Smart Object, and explore how each level of data abstraction affects the user identification accuracy.
