Detecting routes taken by users on public vehicles from their wearables

In this paper, we design a technique to process wearable sensor data in order to detect routes taken by humans as they travel in public buses across the city. Specifically, by processing streams of accelerometer and gyroscope readings from a wearable device worn on the wrist, and then deriving appropriate thresholds, our proposed technique can correctly identify the following activities a) subject boarding a public vehicle; b) starting and stopping of a public vehicle; c) left, right and U-turns of a public vehicle; and d) subject de-boarding a public vehicle. Subsequently, we design an efficient algorithm that takes as an input a) the activities detected above and their corresponding times from the wearable; and b) openly available ground truth data of timings and routes of public buses as they move around the city, and then matches these two to retrieve the possible routes taken by the subject. Using experimental investigations on public transport services offered by two distinct agencies (a city service and a university service), we demonstrate the effectiveness of our proposed techniques to accurately identify routes from processing wearable sensor streams. We are confident that the problem addressed by this paper is unique, and it exposes a unique privacy problem from the perspective of fine grained location tracking of users without their knowledge.