Comparison of different calculations of three-dimensional joint kinematics from video-based system data.

When skin-fixed marker trajectories are used to calculate 3D joint kinematics, the measurement errors (i.e. the difference between the trajectories of the external markers and those of the skeleton) influence to some extent the accuracy of the results, depending both on the calculation method and on the axes about which the rotations are expressed. The purpose of this paper is to compare several expressions of joint angular variations. Two kinematic concepts are used to calculate the changes in the orientation of the distal segment versus the proximal one: the first method consists of computing the components of the spatial attitude vector, the second one deals with the determination of elementary rotations about successive axes. For each of these methods, two sets of three axes are tested to express the results: the axes forming the reference frame affixed to the body segment adjacent to the joint (named fixed axes), and a set consisting of a first axis belonging to the proximal segment, a third axis belonging to the distal segment and a second (floating) axis defined as the cross-product between the two other ones (named mobile axes). To compare these four distinct expressions on the knee joint, numerical simulations of perturbed skin marker trajectories are performed, based on experimental data recorded by a Motion Analysis system during a normal gait cycle. A significant difference is pointed out only for the internal-external rotation angle, for which the best expression - from the viewpoint of sensitivity to experimental errors - is obtained using the components of the attitude vector in a segment-embedded reference frame.