Sharing Life: a Segment-based Semantic Trajectory Annotation Method

With the popularity of GPS equipped smartphone, large volumes of location data associated with mobile objects can be recorded anytime, anywhere. Extracting a user’s behavior patterns embedded helps us to tailor information feeds to better meet a user’s real needs with respect not only to spatio-temporal information but also to previous movement history. This research utilized the raw GPS readings in an attempt to generate semantically meaningful trajectory summaries by modeling individual movement pattern. One supervised movement pattern model for periodic daily routines and the other unsupervised one for all other activities were trained by knowledge-level features including Points of Interest, transportation mode and map-matched route. In addition, a familiarity index of routes was generated representing a user’s personal knowledge about the environment he/she lived in.