This study aims to develop motion trajectory reconstruction algorithms, using captured inertial measurement data from wearable inertial and magnetic sensors mounted near the joints. In this study, we use unscented Kalman filter to derive a quaternion representation of orientation, which describes the coupled nature of orientations in 3-D and is not subject to the problematic singularities associated with an Euler angle representation from an optimal fusion of 9-axis signals, including 3-axis angular rate, 3-axis acceleration, 3-axis magnetic field signals. Then, we can reconstruct movement trajectory by using the proposed upper limb motion model with derived quaternion representation of orientation. In this research, we consider motion acceleration as low-pass filtered noise for compensating disturbance caused by motion. One of the significant contributions of this research is the use of this only one motion acceleration parameter which is modified by cut-off frequency. It can be achieved easily and the reconstructed results are more accurate in comparison to Kalman filter, which learns by trial and error and takes a lot of time. As the verification of developed algorithms, we evaluate the reconstructed trajectory by computing the RMS errors of joint position and correlation coefficients between ideal trajectory and the reconstructed trajectory.
