RAMT

Abstract

Recently a class of new in-vehicle technologies based on off-the-shelf mobile devices have been developed to improve driving safety and experience. For instance, wearables like the smartwatches are utilized to monitor the action of the driver and detect possible secondary tasks. Moreover, wearables can allow a driver to use gesture for in-vehicle controls, reducing distractions to driving. The accuracy of these systems can be significantly improved by tracking the real-time attitude of mobile devices. This paper proposes a novel system called Real-time Attitude and Motion Tracking (RAMT) that can enable a mobile device to accurately learn the coordinate system of a moving vehicle, and hence track its attitude and motion in real time. RAMT consists of a series of lightweight algorithms to sense the vehicle's movement and calculate the device's attitude. It provides a solution for trajectory-based gesture recognition. We have implemented RAMT on a smartphone and a smartwatch and evaluated the performance in 10 real driving trips. Our results show that the overall error of the coordinate system alignment is around 5° for the smartphone and 10° for the smartwatch, and over 84% of customized hand gestures can be accurately recognized with the result of RAMT. A video demo of RAMT is available at https://youtu.be/9rZp7HxyRts.