Effects of Dynamic IMU-to-Segment Misalignment Error on 3-DOF Knee Angle Estimation in Walking and Running

Chao Jiang, Yan Yang, Huayun Mao, Dewei Yang, Wei WangSensors 2022 · DOI

TL;DR

This paper introduces a novel dynamic IMU-to-segment (I2S) alignment method using a modified Particle Swarm Optimization (PSO) algorithm to improve the accuracy of 3-DOF knee angle estimation during walking and running. The study quantifies how different types and magnitudes of I2S misalignment errors affect knee angle estimations across various locomotion speeds, revealing distinct sensitivities for abduction, internal rotation, and flexion.

What Problem Does This Paper Address?

Accurate estimation of joint angles is crucial for biomechanical analysis, particularly for the knee, which is prone to injury. Inertial Measurement Units (IMUs) offer a portable solution for gait analysis, but their accuracy heavily relies on proper alignment between the IMU and the body segment it's attached to (IMU-to-segment, or I2S alignment). Traditional I2S alignment methods are often static, manual, or require complex pre-actions, making them impractical or prone to error. Furthermore, the specific impact of dynamic I2S misalignment errors on 3-DOF knee angle estimation during different activities like walking and running, and how these effects vary across different knee joint movements (abduction, internal rotation, flexion), remains underexplored. This paper aims to address this gap by developing a dynamic I2S alignment technique and systematically investigating the influence of misalignment errors on knee angle estimation across various locomotion speeds.

Methods

The core of this research lies in two main components:

  1. Dynamic I2S Alignment: The authors propose a dynamic weight Particle Swarm Optimization (DPSO) algorithm with a crossover factor, built upon existing "joint constraint" models. This approach uses mathematical constraints derived from the physics of hinge and spherical joints to optimize the I2S alignment parameters in real-time. Unlike static methods, this allows for dynamic adjustments as the subject moves. The DPSO algorithm is enhanced with a dynamic inertia weight and a crossover mechanism to improve convergence speed and avoid local optima, making it more robust for complex motion data.
  2. Quaternion-Based 3-DOF Knee Angle Estimation: After obtaining the dynamic I2S alignment parameters, the IMU data is processed using quaternions to estimate the 3-DOF knee angles (abduction, internal rotation, and flexion). Quaternions are favored for their ability to represent rotations without the singularities associated with Euler angles. The paper details the conversion of rotation matrices (derived from IMU and alignment data) into quaternions and subsequently into the desired Euler angles for the knee joint.

To study the effects of misalignment, the researchers artificially introduced errors (ranging from -10° to +10° in 5° increments) to the dynamically obtained I2S alignment parameters.

Datasets

The study utilized data collected from five healthy participants performing walking (3 km/h), jogging (6 km/h), and ordinary running (9 km/h) on a treadmill. This data was captured simultaneously using:

  • Inertial Measurement Units (IMUs): Xsens Dot wireless nine-axis inertial sensors were attached to the thigh and shank of each leg.
  • Optical Motion Capture System: A twelve-camera high-speed infrared system (Nokov) captured the positions of markers placed on the thigh and shank to serve as the ground truth for knee angle estimation.

Both systems operated at 60 Hz. The collected IMU data was processed using the developed dynamic I2S alignment and quaternion-based estimation algorithms, and then compared against the optical motion capture system data.

Key Insights and Ideas

  • Dynamic Alignment is Key: The paper strongly suggests that static or manual I2S alignment is insufficient for dynamic activities like walking and running. A dynamic approach, as proposed, is crucial for maintaining accuracy.
  • PSO Variants Matter: The proposed DPSO with crossover factor demonstrated superior convergence and accuracy compared to classical PSO and a standard DPSO, highlighting the benefits of adaptive inertia weights and crossover mechanisms for this specific optimization problem.
  • Motion-Specific Sensitivities: The impact of I2S misalignment is not uniform across all knee movements or all locomotion speeds. This implies that a one-size-fits-all approach to error correction or alignment is unlikely to be optimal.
  • Flexion is Robust, Abduction and Internal Rotation are Sensitive: Flexion angles appear to be less affected by I2S misalignment, with errors often behaving as constants that can be corrected. Abduction and internal rotation, however, show greater sensitivity, with their error patterns varying significantly with speed and misalignment type.
  • Thigh vs. Shank Misalignment: The study reveals that IMU-to-thigh misalignment has a more pronounced effect on knee internal rotation estimation, while IMU-to-shank misalignment has the largest numerical impact on knee flexion estimation (though with the smallest standard deviation, indicating consistency).

Key Results

  • Overall Accuracy: The 3-DOF knee angle estimation algorithm achieved Root Mean Square Errors (RMSEs) ranging from 1.6° to 5.9° across different motions when compared to optical motion capture.
  • Effect of IMU-to-Thigh Misalignment:
    • Abduction angle estimation sensitivity decreased from walking to jogging, then slightly increased during running, but remained lower than walking.
  • Internal rotation estimation was more sensitive to IMU-to-thigh misalignment than IMU-to-shank misalignment, with errors increasing from walking to running.
  • Effect of IMU-to-Shank Misalignment:
    • Flexion angle estimation showed the largest numerical error but the smallest standard deviation, suggesting a consistent, constant error.
  • Combined Misalignment: When both thigh and shank had misalignment errors, the estimation error was substantially reduced compared to single-segment misalignment, acting as a control.
  • Motion Differences: Significant differences in RMSEs were observed between walking and running, indicating distinct kinematic differences between these activities.

Limitations and Future Work

  • Error Elimination: The study successfully quantified the effects of dynamic I2S misalignment but did not propose a specific method for error elimination.
  • Future Work: The authors suggest developing models to predict and eliminate the discretization errors caused by different I2S misalignment scenarios to achieve even more accurate joint angle estimations. They also highlight the need for further investigation into the kinematic differences observed during the transition from walking to running.

Summary updated

Abstract

The inertial measurement unit (IMU)-to-segment (I2S) alignment is an important part of IMU-based joint angle estimation, and the accurate estimation of the three degree of freedom (3-DOF) knee angle can provide practical support for the evaluation of motions. In this paper, we introduce a dynamic weight particle swarm optimization (DPSO) algorithm with crossover factor based on the joint constraint to obtain the dynamic alignment vectors of I2S, and use them to perform the quaternion-based 3-DOF knee angle estimation algorithm. The optimization algorithm and the joint angle estimation algorithm were evaluated by comparing with the optical motion capture system. The range of 3-DOF knee angle root mean square errors (RMSEs) is 1.6°-5.9° during different motions. Furthermore, we also set up experiments of human walking (3 km/h), jogging (6 km/h) and ordinary running (9 km/h) to investigate the effects of dynamic I2S misalignment errors on 3-DOF knee angle estimation during different motions by artificially adding errors to I2S alignment parameters. The results showed differences in the effects of I2S misalignment errors on the estimation of knee abduction, internal rotation and flexion, which indicate the differences in knee joint kinematics among different motions. The IMU to thigh misalignment error has the greatest effect on the estimation of knee internal rotation. The effect of IMU to thigh misalignment error on the estimation of knee abduction angle becomes smaller and then larger during the two processes of switching from walking to jogging and then speeding up to ordinary running. The effect of IMU to shank misalignment error on the estimation of knee flexion angle is numerically the largest, while the standard deviation (SD) is the smallest. This study can provide support for future research on the accuracy of 3-DOF knee angle estimation during different motions.
BibTeX
@inproceedings{jiang2022effects,
  author = {Chao Jiang and Yan Yang and Huayun Mao and Dewei Yang and Wei Wang},
  title = {{Effects of Dynamic IMU-to-Segment Misalignment Error on 3-DOF Knee Angle Estimation in Walking and Running}},
  booktitle = {Sensors},
  year = {2022},
  doi = {10.3390/s22229009},
}
Figure