This repository contains materials for the use of IMeasureU inertial measurement units in biomechanics research. These materials were developed as a part of a research study funded by the PAC-12 Student Health and Well-being Grant Program and a collaboration between the University of Colorado, University of Oregon, Stanford University, and University of Southern California. This repository contains code and materials generated during this collaboration and is maintained by Ryan Alcantara. Please open an issue if you have a particular inquiry or need help adapting this approach for your study. The following publications have implemented components of this toolbox:
- "Sacral acceleration can predict whole-body kinetics and stride kinematics across running speeds" by Alcantara et al. (2021)
- "Low-pass filter cutoff frequency affects sacral-mounted inertial measurement unit estimations of peak vertical ground reaction forces and contact time during treadmill running" by Day et al. (2021)
We used 3 IMUs to collect data on the lower legs and sacrum. IMeasureU provides velcro straps to adhere their device just
above the ankle. IMUs on the left/right shank were located laterally on the lower leg with the device's +y axis oriented
proximally and the +z axis oriented medially. Refer to IMU_Blue_Thunder_Specifications.pdf
for more information about the specifications. We developed a custom 3D-printed clip to adhere an IMU to the waistband of
running shorts. Information about the IMU clip can be found in Sacral_IMU_Clip/.
We collected accelerometer, gyroscope, and magnetometer at 500 Hz for all data collections. This was accomplished via the "on board" collection mode, not the "streaming" mode.
IMU and GRF data were collected simultaneously using different software. Additionally, pilot testing revealed discrepancies in temporal synchronization between IMUs. We implemented a 2-step approach to synchronize 1) IMUs to each other, and 2) IMUs to the GRF data. We followed these steps for each data collection:
- Start IMU data collection
- Place all IMUs on swivel desk chair in identical orientation. If you don't have a swivel desk chair, hold all IMUs in your hand.
- Wait 5 seconds to provide a "quiet" period of data collection
- Swivel chair back and forth 3-5 times while pressing down on IMUs to ensure they all rotate together on the chair. If you don't have a swivel desk chair, spin back and forth while holding the IMUs.
- Wait 5 seconds to provide a "quiet" period of data collection
- Place IMUs on participant
- Start GRF data collection via motion capture software
- Instruct participant to stand still on force-measuring treadmill for 5 seconds, perform countermovement jump in-place, and stand still for another 5 seconds
- Start experimental trial(s)
- End experimetnal trial(s)
- Instruct participant to stand still on force-measuring treadmill for 5 seconds, perform countermovement jump in-place, and stand still for another 5 seconds
- End GRF data collection via motion capture software
- Remove IMUs from participant
- Place all IMUs on swivel desk chair in identical orientation
- Wait 5 seconds to provide a "quiet" period of data collection.
- Swivel chair back and forth 3-5 times while pressing down on IMUs to ensure they all rotate together on the chair
- Wait 5 seconds to provide a "quiet" period of data collection
- End IMU data collection
Later on, we'll use the IMU's angular velocity during the swivels to synchronize IMUs to each other and used the IMU's vertical linear acceleration during the countermovement jump to synchronize IMUs to the GRF data. Synchronizing the IMUs before and after GRF data collection allows us to account for any drift in the individual IMU clocks. It is important to note that this approach may generate large file sizes since IMU and GRF data are collected continuously for the entire data collection session. You can modify this approach to fit your needs but ensure that there are opportunities to synchronize IMUs to each other and to GRF data.
When downloading IMU data from the device, ensure that the data collection date and segment name is included in the filename. The scripts in this repository expect slight variations of the following naming conventions for IMU data:
Rshank_2020_10_26.csvLshank_2020_10_26.csvSacrum_2020_10_26.csv
For a list of the acceptable variations for IMU file naming or if you aim to implement additional or different
IMU locations, start by inspecting Preprocessing/support_fxns/importIMUs.m.
Following data collection and data download from IMUs, organize your IMU and GRF data in the following way:
SUBJECT_ID/
- COLLECTION_1/
- COLLECTION_2/
- COLLECTION_3/
- IMU_data/
- Rshank_2020_10_26.csv
- Lshank_2020_10_26.csv
- Sacrum_2020_10_26.csv
- GRF_data.csv
The Baseline_Processing.m script isn't very flexible. It will get mixed up if there are additional CSV/ANC/FORCE files
located in the directory structure above. Our study was longitudinal, so there were multiple collections for each
subject ID.
Next, we will combine IMU csv files and the GRF ASCII file into single MATLAB structure. The scripts we used likely require
modification based on your lab's motion capture software and force-measuring equipment. View Preprocessing/README.md
for information about the script to run and how to modify it for your particular use case. If you get stuck, open an issue
and ask for help!
We developed a standalone MATLAB-based Windows application that temporally synchronizes IMU and GRF data: InSink.
View Synchronization/README.md for more information on how to use the application. If you run into bugs, open an issue!
Now that your IMU and GRF are synchronized, there are so many things you can do! If you're looking for a function that can
quickly identify stance or aerial phase during running, check out the split_steps() function in the Python
or Matlab versions of dryft, an open
source library designed to help process GRF data from force-measuring treadmills.