AN INVESTIGATION INTO CONTEXT-AWARE AUTOMATED SERVICE IN SMART HOME FACILITIES: SEARCH ENGINE AND MACHINE LEARNING WITH SMARTPHONE

Technological advances, in general, coupled with the widespread use of smartphones, create ever more opportunities for mobile applications. This thesis considers the use of such devices within embedded systems to provide automated services in smart home automation. The overall approach links together context-aware data from the physical environment, sensors and actuators for domestic appliances and statistics-based decision-making. A prototype system named ‘Wireless Sensor/Actuator Mobile Computing in the Smart Home’ (WiSAMCinSH) is developed, which in turns aims to provide services that can benefit clients who are currently dependent on others in their daily activities.

This research highlights and covers the following concepts. Firstly, it addresses the need to improve the prototypical decision-making model by enabling it to take into account context-aware information as conditions under which particular action decisions are appropriate. Secondly, an essential aspect of context-aware performance architecture is that its features must be of high accuracy, explicitly readable and fast. Thirdly, it is necessary to determine which probability-based rules are most effective in generating the dynamic environment to control the home facilities. Finally, it is important to analyse and classify in depth the accuracy of context acquisition and the corresponding context control using cross-validation methods.

A case study uses integrated mobile detection technology to improve the efficiency of mobile applications, taking into account the resource limitations forced on the use of mobile devices. It also utilises other embedded sensing technologies to predict expectations, thereby enabling automatic control of facilities in the home. The main approach is to combine search engines and machine learning to create a system architecture for a context-aware computing service. Among the major challenges are finding the best statistics-based rules for decision-making and overcoming the heterogeneous character of the many devices which are used together. The results achieved show very promising potential for the use of mobile applications within a context-aware computing service, albeit one which still presents problems to be resolved through future research.