Wrist tracking for ubiquitous human-machine interface

Abstract

Ubiquitous mobile tracking is a challenging problem especially due to the limitations posed by low-quality sensors available. The noise and bias present in the accelerometers, gyroscopes and magnetometers introduce exponential drift and hence double integration method is insufficient. Sophisticated algorithms have been developed to deal with the challenge, however most of them require expensive offline computations or support from specially installed infrastructure. In this report, we describe a robust system for online mobile tracking with satisfactory accuracy using only a pair of smartphone-watch, called Sound of Motion (SoM). A novel technique of inferring motion state from differential arrival time of sound signal serves as an enabling technology. Three sub-components, namely sound ranging, orientation tracker, and position tracker are discussed in detail. Subsequently, the tracking results of SoM are presented and analysed. Finally, various applications of this system as an input primitive and future improvements are explored.