Smartphone-based vehicular and activity sensing

Abstract

Built-in hardware sensors and network connectivity in many modern smartphones provide opportunities for novel location-based and context-aware applications, particularly in the field of public transportation, physical activity recognition and event detection. A real-time prediction method of the time remaining before reaching the specified destination for bus trips is designed and implemented. The ability to keep track of the trip progress with regards to the list of bus stops to be traversed is built upon the location sensing capability of the passenger's phone and a local database of bus stop positions. This in turn is used in computation of the prediction of remaining time, which involves fetching of historical estimates from a central server and adapting the raw estimate values to the relative speed characteristics of the current trip. Separately, a social sensing and event detection system integrates sound, location and motion sensing data from multiple phones with relevant location-specific data extracted from a Twitter stream as inputs that trigger real-time push notifications back to the phone sensors under specific scenarios. In particular, the accelerometer is used to develop the on-device user activity state recognition logic, for which fairly promising results are obtained.