Locating and identifying user’s meaningful places from raw GPS data

This Master Thesis, in collaboration with Sony, aims to detect meaningful locations in a user's GPS history. Given a stream of GPS recordings (constituted by longitude, latitude, accuracy and timestamp), the goal is to determine what are the most important places in the user’s life. The GPS signal stream is mainly extracted from a user’s smartphone with an average frequency of one position every minute. The GPS signal is subject to errors, mostly when the user steps in a building. The smartphone can also be turned off at any moment so there is also a non-continuous (offline) component. The output is a list of stays with associated characteristics (longitude, latitude, starting time and duration) to be able to label those locations. More precisely, we are mainly interested in recognising home and work.