For more information about Xsens, visit the Xsens Website.

Xsens uses IMU's to track motion, and processes the information in MVN, resulting in an MVNX file or a C3D file. Visual3D can directly read either the C3D or the MVNX file. Please note that raw data from Xsens or any IMU's is not useful for biomechanics without post-processing to generate POSE information.

The legacy implementation of the mvnx model was released in Visual3D v5.02.24. If using Visual3D versions prior to this to process your data, please follow the instructions here. A more flexible implementation of the mvnx model was released in Visual3D 2024.05.2

Typically, Visual3D reads data in the C3D file format which contains target locations from a motion capture system. Although Xsens can export C3D files, there is more information in an MVNX file that Visual3D can use to create a model automatically. Visual3D compatible Xsens .mvnx files should contain the following variables for each segment :
orientation, position, velocity, acceleration, angularVelocity, angularAcceleration. (See “Exporting from MVN Studio” below)

MVNX files can be opened by going to File → Open in Visual3D.

They can also be opened using the File_Open pipeline command.

Model Pose (position and orientation) of the multi-body model

Visual computes the model Pose using a “Link Model” that matches the Xsens link model.

As of Visual3D version 2024.05.2, the user can choose one of two implementations of a model based on the mvnx file.

There are two implementations of a Visual3D model for mvnx data. As users will likely want to use the same implementation each time an mvnx file is loaded, the selection of models can be made in the Program Options, which will save the selection to the registry

When the file is opened, the user will be prompted for all the necessary model metrics to create the model.

The model's root, which is the pelvis, has both an orientation and a position (all 6 degrees of freedom). Visual3D then work through model's chain (assuming a ball and socket at all joints) using the .mvnx segment orienation variables to compute the joint angles. This give us the position and orientation (POSE) of the entire model.

Segment Kinematics:

In the data tree, under KINETIC_KINEMATIC you can see there are “Pose” based variables one could use to reconstruct an the model (ProxEndPos, DistEndPos and CGPos) . Note these Pose based variables are computed directly from .mvnx orientation data (and pelvis position) as described above.

In the KINETIC_KINEMATIC there are also “velocity and acceleration” variables (AngAcc, AngVel, CGAcc, CGVel, DistEndVel and ProxEndVel.) These are obtained from the .mnx file variables : velocity, accelaration, angularVelocity and angularAcceleration. (The velocity and acceleration are at the segment origin.) Note these variables and not are not obtained via derivatives of the pose variables. (Relative motion equations are used to compute the “velocity and acceleration” variables from the Xsens input.)

Finally for the other kinematic variables that can be obtained via Compute_Model_Based_Data, position and orientation based variables come from the “Pose” variables in the KINETIC_KINEMATIC tree and velocity and acceleration data come from “velocity and acceleration” variables in the KINETIC_KINEMATIC tree.  

Segment Kinetics:

Once Visual3D creates the KINETIC_KINEMATIC variables: ProxEndPos, DistEndPos, CGPos, AngAcc, AngVel, CGAcc (and the ground reaction force if needed) we have all we need to compute the inverse dynamics.

Note:

If you take the first derivative of the ProxEndPos KINETIC_KINEMATIC variable it will not exactly equal the ProxEndVel KINETIC_KINEMATIC variable which come from the .mvnx Translational Velocity of the Origin. The idea is to use the Xsens data for velocity, acceleration and kinetic calculations even if it is slightly inconsistent with the derivative of the Pose data. (This data is assumed to be more accurate measure of velocity and acceleration then taking the derivatives of which amplifies noise.)

Visual3Dv2020.## supports reading MVNX files with one prop sensor. The sensor will be loaded automatically as a segment (ex. PR1). If more props are required to process your data, please contact C-Motion. When using an Xsens prop in Visual3D, all information is loaded automatically and no further steps are required. A sample MVNX file with a prop can be downloaded here.

This tutorial explains how to adjust the location of the prop segment and assign a graphics object to the prop using a script.

Unlike the legacy implementation when the file is opened, the user will not be prompted for the anthropometric data (i.e. mass and moment of inertia). Instead default values are used, which can later be modified by the user. Our thinking was that users that are only interested in kinematics do not want to be prompted for unnecessary information every time a file is loaded.

In this implementation, the information in the mvnx file is used to create 4×4 ROTATION signals similar to the output from markerless tracking systems like Theia3D, and Kinatrax, and is consistent with the ROTATION signals created from bvh files (Fujitsu implementation of bvh).

Using model based information in the mvnx file, a model is created automatically from the ROTATION signals.

In this implementation the Kinetic_Kinematic variables are created consistently with the other ROTATION based models.

Unlike the legacy implementation the segment pose is explicitly defined.

If a prop sensor exists in the mvnx file, a ROTATION matrix is created form the prop pose.

The user can create/modify a segment definition using this pose

• Sample MVNX files
  • These sample MVN files were downloaded from the Xsens Website and then exported from MVN Studio as MVNX files

When exporting the mvnx file from MVNX studio, it is important to export the Position, Orientation, Velocity, Angular Velocity, Acceleration, and Angular Acceleration should be exported. Below are instructions to complete this in MVN Studio.

Exporting from MVN Studio (click to expand)

2. Modify the export settings - Make sure the following are selected: - Position - Orientation - Velocity - Angular Velocity - Acceleration - Angular Acceleration If any of these signals (translational velocity, translational acceleration, angular velocity, angular acceleration) are missing from the MVNX file, the derivatives of the position data will be used to calculate these signals. NOTE: Once you have set this, these changes will be saved even after MVN Studio is closed. So you only need to do this step once.

3. Load the data