Context awareness for wearable computers
The research described in this thesis considers mobile technology with particular reference to the use of context sensing. It is argued that such technologies are useful to enhance user proficiency in everyday tasks. A wearable, context–aware computer system and a set of evaluation tasks are devised to investigate this premise. A photograph diary study is carried out to elicit defining features of a broad range of everyday activities. These features are called Context Identifiers. From this a structured definition of context is suggested that bridges the gap between the current theoretical and technological definitions of context. Based on a literature review of current mobile and wearable technology and on the findings of the photograph diary study, a novel wearable computer and supporting software is developed. The wearable computer can detect and interpret features of everyday context, including Location, Posture and Movement, and Objects. During the design cycle of the wearable computer, experiments are conducted to evaluate three versions of the wearable computer. The usability of the computer is considered based on measures of efficiency, effectiveness and user satisfaction. Use of the system is shown to improve user task proficiency in the completion of simple tasks and the wearable computer is shown to capture context in a similar way to humans. Specifically when the first version of the system is used in an information retrieval task, the wearable computer is shown to significantly decrease task completion time when compared to using a standard internet enabled computer or users searching for information in their environment. In addition the task accuracy is increased. The second version of the system in which the number of everyday contextual features the system can detect is increased, again significantly decrease the task completion time when compared to the same system detecting less context features. The third version of the system which detects further more contextual features is shown to be highly usable based on a number of usability measures and is shown to capture context in a similar fashion to humans.