Inertial Compensation for Platform Movement of an Instrumented Treadmill

This is the source code which reproduces the results for the paper entitled:

""

Instrumented treadmills provide a convenient means for inverse dynamic gait analysis.
Recently, these instrumented treadmills have the capability of translating and rotating the walking surface. Load cells, located directly underneath the force plate, incorrectly add the inertia of the moving mass of the treadmill to the measured ground reaction forces (GRF) of the test subject. This codes provides a computation method for reducing these inertial errors and produces all results shown in the above publication.

Dependencies

• Matlab R2014b (8.3.0.532)
• Signal Processing Toolbox (Version 6.21)

Basic Installation

Download the source code and data from the Git repository:

$ wget https://github.com/csu-hmc/inertial-compensation/archive/master.zip
$ unzip inertial-compensation-master.zip
$ cd pitch-moment-compensation-master

Usage

The complete results can be computed by starting Matlab with the main source code directory in your Matlab path and typing the following at the command prompt:

>> inertial_compensation_test

Once the function has run you will find all the plots in the Results directory. For the cutoff frequency of interest, a table of results including the reduction in the root-mean-square (RMS) between uncompensated and compensated force (Fxyz,Mxyz) and the percent difference between the them

Computation Steps

The data processing pipeline follows this general process:

1. Specifies filenames located in the Data directory.
2. Parses pertinent data from tab separated ACSII data files (data_parser.m). Data includes the timestamps, recorded 3D coordinates of reference markers located along the treadmill frame, recorded 3D force plate data from two force plates, and recorded 3D accelerations of 4 accelerometers located on each rigid corner of the treadmill
3. Sets a range of desired cutoff frequencies for the lowpass filter (1-20 Hz). Also sets a cutoff frequency of interest (6 Hz).
4. Filters signals (markers,forces, accelerations) with the cutoff frequency. Truncates the first and last second of all data
5. Adds signal delay to the force and acceleration signals (add_delay.m). Truncates any NaN from all data
6. Computes a matrix of calibration coefficients through linear least squares between the measured accelerations (input) and force signals (output).
   (compensation.m)
7. Multiplies the recorded accelerations from another trial with the calibration coefficients. Subtracts the result from the recorded force signals (compensation.m)
8. Performs a coordinate transformation on the compensated forces ('transformation.m') using the Soderquist method (soder.m). The speed soder method is sincreased through multidimensional matrix multiplication (mmat.m)
9. Plots the uncompensated vs. compensated force signals (Fx,Fy,Fz,Mx,My,Mz) for force plate 1 (plot_compensation_graph.m)
10. Performs the inverse dynamic analysis on recorded walking data. Adds uncompensated and compensated force signals to the walking data to see the affect on hip, knee, and ankle joint torques. Plots the results (sensitivity_analysis.m)
11. Computes the RMS of the uncompensated and compensated force signals, as well as the percent reduction. Displays a table of results for the desired frequency (6 Hz) (calculate_statistics.m)
12. Shows a comparison of RMS with respect to the filter frequency for compensated and uncompensated force and moment signals (Fxyz, Mxyz). If the range of the cutoff frequencies is 1, then the graph will not generate. (plot_frequency_graphs.m)
13. Saves the graphs to the Results directory.
14. Performs validation test calculations (validation_test.m), using 10 trials of 5 different treadmill movements. Coefficients are determined from one trial from each of the 5 movements. These coefficients then reduce inertial errors in the remaining trials of the same movement and the other movements.
15. Calculates the mean and standard deviation of the compensated RMS values for all of the validation tests and saves into an Excel file in the Results directory.