gait_signatures

This is code related to the paper: Taniel S. Winner, Michael C. Rosenberg, Kanishk Jain, Trisha M. Kesar, Lena H. Ting, Gordon J. Berman. Discovering individual-specific ‘gait signatures’ from data-driven models of neuromechanical dynamics. Plos Comp Biology. 2023;

Data

PareticvsNonP_RNNData.csv: Data set containing sagittal plane joint angles of 5 able-bodied and 7 stroke survivors as they walked on a treadmill at 6 different speeds each for 15 seconds. (72 trials, 1500 sample/trial). The dataset was collected at Emory Rehab Hospital Motion Analyses Lab (PI Kesar).

Subjectslabels.csv: Data set containing a list of subject identifiers corresponding to each trial.

Speedlabels.mat: Data set containing a list of treadmill speeds (cm/s) corresponding to each trial.

Python Code

Please ensure that you are using the specific versions of tensorflow and keras below by running the following commands in script

!pip install keras==2.15.0

tensorflow==2.15.0

Versions for other packages used in this code base

• Scikit-learn: 1.2.2
• Matplotlib: 3.7.1
• Pandas: 1.5.3
• NumPy: 1.25.2
• SciPy: 1.11.4
• CSV: 1.0

Gait Signatures Script 1: Train Model Architectures.ipynb

Gait Signatures Script 2: Extract Internal Parameters.ipynb

Gait Signatures Script 3: Generate Gait Signatures.ipynb

Gait Signatures Script 4: Short-TimeSelf-Driven Signatures.ipynb

Gait Signatures Script 5: Compute Gait Signature Alignment.ipynb

Gait Signatures Script 6: Compare Gait Signature Alignment.ipynb

Gait Signatures Script 7: Save Hyperparameter Results Single Folder.ipynb

Code was executed on Google Colab. The specific Collab software versions tested are Python 3.8.16 and Tensorflow 2.9.2. on Google Colab’s standard GPU with high-RAM runtime (54.8 gigabytes).

Mathematical tools

We make use of a collection of mathematical tools used in the Biologically Inspired Robotics and Dynamical Systems (BIRDS) lab at the University of Michigan. Most of these tools were either coded or invented to a large degree by the lab's founder, Shai Revzen.

fourierseries.py: a general purpose FourierSeries class

phaser.py: the Phaser algorithm from doi:10.1103/PhysRevE.78.051907

jac.py: Jacobian functor (maps a function to its Jacobian)

odedp5.py: order 5 ODE integrator with special facilities for hybrid systems

sde.py: order 3 Stratonovic SDE integrator

fastode: aggressively optimized version of the odedp5 integrator for very fast operation. Requires inner ODE function to be in C

util.py, shaiutil.py and shai_util.py: helper functions