A low-cost markerless motion capture system to automate functional gait assessment: Feasibility study

Functional gait assessments in older adults have traditionally required manual in-person quantification of clinical measures such as walking speed and step placement. This reliance on individuals trained in motion analysis hinders the frequency with which they are performed, and reduces their generalizability and replicability. To standardize, simplify, and broaden access to gait assessments we here deploy recently-developed open-source tools to produce a low-cost, AI driven markerless motion capture system with custom analysis software for Functional Gait Assessment. Our system uses 3 Cameras, and was validated with a traditional marker-based system on subjects (N = 3), showing strong correlation to laboratory-standard measures of step length (R2 = 0.98), step width (R2 = 0.97), and head speed (R2 = 0.95). The markerless system’s FGA reports demonstrated data similar to previous FGA findings in older adult subjects (N = 5). Moreover, supplemental standard biomechanical gait measures St..., , # Data from: A low-cost markerless motion capture system to automate functional gait assessment: Feasibility study Dataset DOI: [10.5061/dryad.2bvq83c2t](10.5061/dryad.2bvq83c2t) ## Description of the data and file structure Jeremy D. Wong, William J. Herspiegel, and Arthur D. Kuo ## Quickstart After downloading files, clone the github repository [kinarmdataanalysis](https://bitbucket.org/jdwongmcl/kinarmdataanalysis/src/master/), and run files prefixed with 'gofigures_psfmc_swsl' to regenerate figures testing step length and width hypothesis testing from the manuscript. ### Files and variables #### File: individual subject files for each subject in experiment 1. #### File: jer_psfmc.mat **Description:** phasespace-freemocap data: ##### Variables: 3D trajectories for the following * *fmc_bodies:* (COCO (Common Objects in Context) Keypoints Dataset, 17 keypoints) for each of 6 trials (regular, wide): extracted gait only * *fmc_bodies*: (COCO (Common Objects in Context) Keypoin..., I received explicit consent for de-identified data in the public domain. We used randomized unique string identifiers for each subject. ,