Personal sensing from mobile devices: the feasibility of mobile devices for diverse applications

Mobile phones are the most plentiful, always on, and connected computational platform in existence. Mobile phones are also the most ubiquitous sensing platform. The ability of these devices to sense and infer information about the world around them enables us to create—heretofore impossible—timely, contextual, and relevant systems. The world is faced with a number of daunting problems to solve; growing energy crises, global warming, and crippling obesity in the face of world hunger are all very real problems in our daily lives. More powerful computers will allow experts to study these problems at a large scale; however, these problems remain very real and present; many require more than simple policy changes or guidelines; indeed they require the distributed effort of hundreds of thousands of individuals, each of whom is already overwhelmed with too many things to keep track of and too many tasks vying for time. The goal of this dissertation is to build practical activity recognition systems, initially through building custom hardware platforms like the Mobile Sensing Platform, and later through the use of sensor enabled mobile phones; to enable a wide range of systems to more intelligently combat societal problems.