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

Area of Science:

• Biomechanics
• Kinetics and Kinematics
• Sensor Technology

Background:

• Accurate 3D joint angle measurement is crucial for biomechanical analysis.
• Inertial sensors offer a portable solution but their accuracy requires rigorous validation.
• Previous studies often confounded sensor error with sensor attachment error.

Purpose of the Study:

• To quantify the intrinsic accuracy of inertial sensors for 3D knee joint angle measurement.
• To isolate sensor error from sensor attachment error using a novel gimbal apparatus.
• To establish a benchmark for inertial sensor performance in biomechanics.

Main Methods:

• Utilized an instrumented gimbal rotating about three axes to provide ground truth joint angles.
• Employed sensor attachment devices fixed to the gimbal to eliminate attachment-related errors.
• Calculated angle estimation errors (RMSE, Bland-Altman) for inertial sensors in flexion/extension, abduction/adduction, and internal/external rotation.

Main Results:

• Root Mean Square Error (RMSE) for inertial sensors: 3.20° (flexion/extension), 3.42° (abduction/adduction), 2.88° (internal/external rotation).
• Bland-Altman means of maximum absolute error: -1.63° (flexion/extension), 3.22° (abduction/adduction), -2.61° (internal/external rotation).
• Errors are comparable to or smaller than those reported in previous human gait studies.

Conclusions:

• Inertial sensor-based 3D joint angle measurement inherently involves errors of a few degrees.
• Sensor attachment significantly contributes to error in previous human gait studies.
• The developed apparatus and methodology can validate inertial sensor systems and algorithms.