Smartphone sensing.
Emiliano Miluzzo · 2011
The increasing popularity of smartphones with their embedded capability and the availability of new application distribution channels, such as, the Apple AppStore and the Google Android Market, is giving researchers a unique opportunity to deploy mobile applications at unprecedented scale and collect sensor data way beyond the boundaries of traditional small-scale research laboratory deployments. This thesis makes a number of contributions to smartphone by introducing new models, algorithms, applications, and systems. First, we propose CenceMe, the first large-scale personal and social application for smartphones, which allows users to share their real-time sensing presence (i.e., activity and context) with friends using the phone, web, and social network sites (i.e., Facebook, Myspace, Twitter). CenceMe exploits the smartphone's onboard sensors (viz. accelerometer, microphone, GPS, Bluetooth, WiFi, camera) and lightweight, efficient machine learning algorithms on the phone and backend servers to automatically infer people's activity and social context (e.g., having a conversation, in a meeting, at a party). The development, deployment, and evaluation of CenceMe opened up new problems also studied in this dissertation. Sensing with smartphones presents several technical challenges that need to be surmounted; for example, the smartphone's context (i.e., the position of the phone relative to the event being sensed varies over time) and limited computational resources present important challenges that limit the inference accuracy using phones. To address these challenges, we propose an evolve-pool-collaborate model that allows smartphones to automatically adapt to new environments and conduct collaborative among co-located phones resulting in increased robustness and classification accuracy of smartphone in the wild. We call this system, Darwin Phones. The final contribution of this dissertation explores a new mobile application called VibN, which continuously runs on smartphones allowing users to view live feeds associated with hotspots in a city; that is, what is going on at different locations, the number of people and demographics, and the context of a particular place. VibN addresses a number of critical problems to the success of smartphone sensing, such as, running continuous algorithms on resource limited smartphones, resolving privacy issues, and developing a sensor data validation methodology for applications released via the app stores (i.e., validating sensor data and identifying patterns without any notion of ground truth evidence). Such a methodology is crucial to the large-scale adoption of smartphone in the future. Smartphone is an emerging field that requires significant advances in mobile computing, machine learning, and systems design. It is an exciting area of research that is cross-disciplinary and likely to touch on many application areas and scientific domains moving forward. The work presented in this dissertation identifies new problems and solutions that help advance our understanding in what is now a fast-moving area of research.