Quantification of inertial sensor-based 3D joint angle measurement accuracy using an instrumented gimbal

This study quantified the accuracy of inertial sensors in 3D anatomical joint angle measurement with respect to an instrumented gimbal. The gimbal rotated about three axes and directly measured the angles in the ISB recommended knee joint coordinate system. Through the use of sensor attachment devices physically fixed to the gimbal, the joint angle estimation error due to sensor attachment (the inaccuracy of the sensor attachment matrix) was essentially eliminated, leaving only error due to the inertial sensors. The angle estimation error (RMSE) corresponding to the sensor was found to be 3.20° in flexion/extension, 3.42° in abduction/adduction and 2.88° in internal/external rotation. Bland-Altman means of maximum absolute value were -1.63° inflexion/extension, 3.22° in abduction/adduction and -2.61° in internal/external rotation. The magnitude of the errors reported in this study imply that even under ideal conditions irreproducible in human gait studies, inertial angle measurement will be subject to errors of a few degrees. Conversely, the reported errors are smaller than those reported previously in human gait studies, which suggest that the sensor attachment is also significant source of error in inertial gait measurement. The proposed apparatus and methodology could be used to quantify the performance of different sensor systems and orientation estimation algorithms, and to verify experimental protocols before human experimentation.