Journal: bioRxiv

Article Title: An open-source and wearable system for measuring 3D human motion in real-time

doi: 10.1101/2021.03.24.436725

Figure Lengend Snippet: The components and default IMU orientations of the OpenSenseRT System. (A) An IMU on the pelvis is required and acts as the base in order to compute the relative orientation of other sensors. The OpenSenseRT System accommodates a variable number of additional IMUs to customize which kinematics are measured. To monitor movement of the upper body, three IMUs may be placed on each arm (on the upper arm, forearm, and hand). An additional IMU can be placed on the torso. The orientation frame with axes shown in red, green, and blue is used to orient the x, y , and z axes defined on each IMU. These individual body frames should align with the world reference frames of the fore-aft, mediolateral, and vertical axes, while the subject’s joint segments are aligned in a neutral standing (or other known) position. (B) The lower-limb IMU placements also require the pelvis IMU as a base and can include up to three IMUs on the thigh, shank, and foot of each leg. OpenSenseRT allows for 1 to 14 of these IMUs to be sampled in a custom configuration. (C) A zoomed in view of the system components shows the microcontroller, battery, button (for starting and stopping recordings), IMU connector, and pelvis IMU.

Article Snippet: The second thread uses these orientations and the musculoskeletal model as inputs to the inverse kinematics solver included in OpenSim version 4.2 [ ] to compute the generalized coordinates of the model (i.e., joint angles and joint angular velocities).

Techniques: Battery

Journal: bioRxiv

Article Title: An open-source and wearable system for measuring 3D human motion in real-time

doi: 10.1101/2021.03.24.436725

Figure Lengend Snippet: Flowchart for computing joint kinematics. The user defines the set of body segments to track, the IMU ports associated with the segments, and the computation frequency. The calibration process consists of a heading correction for the given IMUs and initializing the pose of the musculoskeletal model. The calibrated model and initial pose are passed to the inverse kinematics solver. One thread on the microcontroller records raw IMU data and computes orientations for each body segment using a Mahony Filter at each step. A second thread takes these orientations and the musculoskeletal model to solve the inverse kinematics, estimating joint kinematics at each time step.

Article Snippet: The second thread uses these orientations and the musculoskeletal model as inputs to the inverse kinematics solver included in OpenSim version 4.2 [ ] to compute the generalized coordinates of the model (i.e., joint angles and joint angular velocities).

Techniques:

Journal: bioRxiv

Article Title: An open-source and wearable system for measuring 3D human motion in real-time

doi: 10.1101/2021.03.24.436725

Figure Lengend Snippet: Lower-limb joint kinematics computed with the OpenSenseRT system compared to optical motion capture. (A) Joint kinematics during walking at 1.25 m/s were computed in real-time at 30 Hz with the IMU system for five subjects. The error bands represent one standard deviation of the RMSE across all subjects. (B) Joint kinematics during running at 4.0 m/s were computed offline (represented with the dash-dot line) at 100 Hz with the IMU system for a single subject. The error bands represent one standard deviation of the RMSE across all gait cycles for the single subject. Running requires a faster computation rate than the IMU system can perform consistently in real-time, although it can be analyzed in real-time over short bursts by using all cores.

Article Snippet: The second thread uses these orientations and the musculoskeletal model as inputs to the inverse kinematics solver included in OpenSim version 4.2 [ ] to compute the generalized coordinates of the model (i.e., joint angles and joint angular velocities).

Techniques: Standard Deviation

Journal: bioRxiv

Article Title: An open-source and wearable system for measuring 3D human motion in real-time

doi: 10.1101/2021.03.24.436725

Figure Lengend Snippet: RMSEs for joint kinematics computed with the IMU system compared to optical motion capture over time for walking. The RMSE increased from approximately 3.5 to 6 degrees over the first 150 seconds of the condition. Hip rotation had the largest RMSE and drift. The black line and grey error band represent the average and standard deviation in RMSE across all subjects and joints.

Article Snippet: The second thread uses these orientations and the musculoskeletal model as inputs to the inverse kinematics solver included in OpenSim version 4.2 [ ] to compute the generalized coordinates of the model (i.e., joint angles and joint angular velocities).

Techniques: Standard Deviation

Journal: bioRxiv

Article Title: An open-source and wearable system for measuring 3D human motion in real-time

doi: 10.1101/2021.03.24.436725

Figure Lengend Snippet: Upper-limb joint kinematics computed with the IMU system compared to optical motion capture. (A) An upper-limb task from the Fugl-Meyer assessment simulated cutting food. Subjects picked up a cutting utensil bent over a dish, and performed a cutting motion before placing the utensil back down. The task was repeated ten times to emulate an upper-limb “activity cycle” to look at average kinematics for one subject. (B) In the trunk range of motion task, subjects performed trunk flexion, bending, and then rotation. In both (A) and (B), the bands represent one standard deviation of the RMSE across all subjects.

Article Snippet: The second thread uses these orientations and the musculoskeletal model as inputs to the inverse kinematics solver included in OpenSim version 4.2 [ ] to compute the generalized coordinates of the model (i.e., joint angles and joint angular velocities).

Techniques: Activity Assay, Standard Deviation