Hierarchical MARG Fusion Method With Human Constraints for Upper Limb Motion Estimation

Motion capture (MoCap) provides technical support for human motion tracking and orientation estimation, which has enormously benefited the research of human–computer interaction in the aerospace field. Compared with the optical MoCap system, inertial measurement units (IMUs) play an irreplaceable role in accurately calculating the orientation and position of the human body without being affected by light. This article presents a novel approach that incorporates the coupling of position and orientation into the state equation, enabling accurate estimation of human motion. Also, we incorporate constraints into the optimization process using a least squares-based projection method to reduce drift and improve long-term stability. First, on the basis of the established upper limb kinematics model, the quaternion-based IMU fusion extended Kalman filter (EKF) method is leveraged to compensate roll and pitch angles. Then, the magnetic, angular rate, and gravity (MARG)-EKF method performs a complete compensation of 3-D orientation and position through the data fusion of MARG. In order to further reduce the drift, human constraints are introduced. The novel constrained MARG-EKF algorithm has been tested by various movements, which shows the position tracking accuracy in
${X}$
-,
${Y}$
-, and
${Z}$
-axes of 10 cm. Also, the imposed constraints can reduce the error by 2.4%–15.0% for the forearm and 3.2%–11.4% for the upper arm. The experimental results represent that our proposed method can track upper limb motion more accurately with higher feasibility and stability.