Abstract
The investigation of rotation behavior in human beings enjoys a longstanding and enduring interest in laterality research. While in animal studies the issue of accurately measuring the number of rotations has been solved and is widely applied in practice, it is still challenging to assess the rotation behavior of humans in daily life. We propose a robust method to assess human rotation behavior based on recordings from a miniature inertial measurement unit that can be worn unobtrusively on a belt. We investigate the effect of different combinations of low-cost sensors-including accelerometers, gyroscopes, and magnetometers-on rotation measurement accuracy, propose a simple calibration procedure, and validate the method on data from a predefined path through and around buildings. Results suggest that a rotation estimation based on the fusion of accelerometer, gyroscope, and magnetometer measurements outperforms methods based solely on earth magnetic field measurements, as proposed in previous studies, by a drop in error rate of up to 32 %. We further show that magnetometer signals do not significantly contribute to measurement accuracy in short-term measurements, and could thus be omitted for improved robustness in environments with magnetic field disturbances. Results also suggest that our simple calibration procedure can compete with more complex approaches and reduce the error rate of the proposed algorithm by up to 38 %.
Leuenberger, K