Adaptive Kalman filter for orientation estimation in micro-sensor motion capture

One of the biggest challenges in micro-sensor motion capture is the drift problem caused by integration of angular rates to obtain orientation estimation. To reduce the drift, existing algorithms make use of gravity and earth magnetic field measured by accelerometers and magnetometers. Unfortunately, the gravity measurement can be strongly interfered by human motion acceleration. This paper presents a quaternion-based adaptive Kalman filter for drift-free orientation estimation. In this filter, the motion acceleration associated with the quaternion is included in the state vector, to compensate the effects which human body motion may have on the reliability of gravity measurement. Moreover, the process noise covariance is adapted based on the variation of sensor signals, to optimize the performance under human motion existence. The final experiments show that the proposed algorithm has least error compared with the existing methods, and the use of motion acceleration compensation together with the adaptive mechanism can improve the accuracy of human motion capture.