Abstract： Soft tissue artifact (STA) is the main error source in upper extremity motion tracking using skinmounted marker sets, especially in upper arm axial rotation. To address this problem, an online upper arm axial rotation angular tracking error compensation method was proposed using matrix decomposition of the relative pose data between the connected bodies based on the joint DOFs constraint. Frist, the transformation matrix between the coordinate of each limb segment to the responding tracking sensor was calculated by optimization method using a group of upper extremity motion information with no upper arm axial rotation. In this situation, the relative attitude from the forearm to the upper arm would comply with the two DOFs constraint of elbow. Then, an error compensation motion for the tracking deviation caused by the STA of upper arm axial rotation was introduced into the relative attitude. The compensation angle was calculated online by the decomposition of this relative attitude matrix using three DOFs rotational motions when the motion tracking of the movement of upper extremity with upper arm axial rotation. Vivo upper extremity motion tracking experiments of four subjects show that this method can effectively online fix the upper arm axial rotation tracking error caused by the STA. The compensation accuracy is better than that of the linear error compensation method, especially when the rotation angle range is large.