This paper presents a novel approach for mobile phone centric observation of a user’s
facing direction, relying solely on built-in accelerometer and magnetometer. Our approach
achieves greater accuracy and independence by an automatic detection of the wearing
position of the mobile device on the user’s body and subsequent selection of an optimum
strategy for estimating the user direction. We report on a detailed analysis of various
features sets and classifiers in order to determine an optimum selection of those for
recognizing the wearing position and evaluate the accuracy and reliability of our overall
implementation through an extensive measurement campaig