Smoother-based 3D Foot Trajectory Estimation Using Inertial Sensors.

OBJECTIVE: Measuring 3D foot trajectories with foot- worn IMUs (inertial measurement units) is essential for a variety of applications, such as gait analysis and fall risk assessment. IMU-based foot trajectory is reconstructed usually by double integrating the global coordinate acceleration, in which drifts of signals are accumulated and lead to unbounded error increase. To reduce drift errors, a smoother-based method is proposed in this paper.

METHODS: The smoother-based method not only corrects initial values of integrations, but also smooths integrating processes through a backward update. Both the orientation estimation and velocity estimation are improved in this concept, which contribute to the improvement of the trajectory estimation.

RESULTS: The final results are compared with an optical motion capture system as references. Accuracy is evaluated with ground level-walking of nine adult participants, 2302 strides in total. Errors are reduced by 62% on stride length and 44% on stride width of the estimation without our method, with final errors -0.24 ± 1.11 cm and -0.02 ± 0.95 cm.

CONCLUSION/SIGNIFICANCE: Results prove that our method can improve the accuracy of 3D foot trajectory estimation. Furthermore, this smoother-based method can reduce drift-related errors when estimating trajectories, which will allow it to expand its applications into other IMU-based measurements.