cs.CVApr 13, 2026

Fall Risk and Gait Analysis in Community-Dwelling Older Adults using World-Spaced 3D Human Mesh Recovery

Authors: Chitra Banarjee, Patrick Kwon, Ania Lipat, Rui Xie, Chen Chen, Ladda Thiamwong

Organizations: College of Medicine, UCF · Institute of AI, UCF · College of Nursing, UCF · College of Sciences, UCF

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Gait assessment is a key clinical indicator of fall risk and overall health in older adults. However, standard clinical practice is largely limited to stopwatch-measured gait speed. We present a pipeline that leverages a 3D Human Mesh Recovery (HMR) model to extract gait parameters from recordings of older adults completing the Timed Up and Go (TUG) test. From videos recorded across different community centers, we extract and analyze spatiotemporal gait parameters, including step time, sit-to-stand duration, and step length. We found that video-derived step time was significantly correlated with IMU-based insole measurements. Using linear mixed effects models, we confirmed that shorter, more variable step lengths and longer sit-to-stand durations were predicted by higher self-rated fall risk and fear of falling. These findings demonstrate that our pipeline can enable accessible and ecologically valid gait analysis in community settings.

Figures & tables

Figure 1: Schematic of Timed Up and Go (TUG) test set up
VariableAssessmentSignificance/Interpretation
Self-Rated Fall RiskCenters for Disease Control STEADI Score12-item questionnaire quantifying self-rated fall risk, summed to produce the STEADI score, with higher scores indicating higher risk. Validated as indicator of fall risk [ 4 , 43 ] .
Fear of Falling (FoF)Short Falls Efficacy Scale-International (FES-I) Score7-item questionnaire evaluating FoF, with higher scores indicating higher FoF. Validated as indicator of fall risk [ 13 , 25 , 63 ] .
Postural SwayBTrackS Balance Test ScoreStatic balance assessment in which participants stand on a balance plate for 20 seconds three times to evaluate postural sway, with higher measures indicating worse static balance. Validated as indicator of fall risk [ 31 , 45 , 19 ] .
Table 1: The measures in this table were completed by the participants to provide information about subjective and objective fall risk, specifically self-rated fall risk, fear of falling, and postural sway.
Figure 2: The overall pipeline of our study. Basing on the world trajectory extraction results from the TUG test using GVHMR [ 53 ] (Sec. 3.2), we analyze the gait factors (Sec. 3.3) and conduct statistical analysis (Sec. 3.4).
ComponentIndicator
y˙hip\dot{y}_{\mathrm{hip}}Vertical hip trajectory indicates initiation of STS, weighted the highest.
z˙shoulder\dot{z}_{\mathrm{shoulder}}Anterior translation of the upper trunk during forward flexion; reliably marks the preparatory lean phase that precedes vertical rise.
θ˙trunk\dot{\theta}_{\mathrm{trunk}}Angular trunk flexion–extension is noisier and more sensitive to small movements, thus assigned a lower weight.
Hip Line SignalAlternating signed extrema in hip-line velocity reflect rapid angular deceleration and reversal of pelvic rotation during turning.
Table 2: Components of signals for segmentation of TUG subtask (sit-to-stand, turns, stand-to-sit).
Figure 3: Automated Subtask Segmentation for three trials. Solid line refers to the hip line velocity and dashed line refers to sit-to-stand and stand-to-sit (STS) signal.
 STEADI Score
 Estimate (CI) pp
STS Duration1.231.23 (0.45,2.01)(0.45,2.01) 0.002\mathbf{0.002}
Step Length (SL)−1.36-1.36 (−2.03,−0.68)(-2.03,-0.68) <0.001\mathbf{<0.001}
SL Variability−19.62-19.62 (−30.44,−8.80)(-30.44,-8.80) <0.001\mathbf{<0.001}
 Short FES-I Score
 Estimate (CI) pp
Table 3: Linear mixed effect model results of STEADI Score (self-rated fall risk) and short FES-I (fear of falling). ST refers to the video-derived step time. Estimate refers to the β\beta coefficient, CI = confidence interval, pp = pp -value.
Figure 4: Scatterplots of mean step length (SL), SL variability, and sit-to-stand and stand-to-sit (STS) duration with fall risk factors (STEADI: self-rated fall risk, Short FES-I: fear of falling).
VariableMean ±\pm Standard Deviation
Age74.2±7.8174.2\pm 7.81 years
STEADI Score2.95±2.832.95\pm 2.83
Short FES-I Score9.40±3.029.40\pm 3.02
BTrackS Balance Score29.3±11.229.3\pm 11.2 cm
Table 4: Population characteristics.
PredictorEstimateCIp
(Intercept)296.47[160.02, 432.92]<0.001\mathbf{<0.001}
ste8-0.14[-5.34, 5.05]0.956
age-2.79[-4.73, 0.86]0.005\mathbf{0.005}
Table 5: Step Width Variability
Figure 5: XSENSOR Insole and IMU System
Figure 6: Snapshots of research assistant completing the TUG (left), with the camera-centric view overlay of the HMR (middle), and the global-centric view (right).
Figure 7: Detection of heel strike using peak detection, according to the detected joints using GVHMR.
 Model 1Model 2Model 3
PredictorsEstimatesCIpEstimatesCIpEstimatesCIp
(Intercept)49.74[32.01, 67.46]<0.001\mathbf{<0.001}65.08[48.45, 81.72]<0.001\mathbf{<0.001}65.69[47.95, 83.44]<0.001\mathbf{<0.001}
STEADI-1.36[-2.03, -0.68]<0.001\mathbf{<0.001}——————
Age-0.00[-0.25, 0.25]0.988-0.14[-0.37, 0.10]0.264-0.23[-0.48, -0.01]0.065
Short FES-I———-1.04[-1.65, -0.43]0.001\mathbf{0.001}———
BTrackS——————-0.11[-0.27, -0.05]0.176
Table 6: Step Length (SL) Analysis
 Model 1Model 2Model 3
PredictorsEstimatesCIpEstimatesCIpEstimatesCIp
(Intercept)296.47[160.02, 432.92]<0.001\mathbf{<0.001}301.74[182.35, 421.13]<0.001\mathbf{<0.001}302.56[182.97, 422.14]<0.001\mathbf{<0.001}
STEADI-0.14[-5.34, 5.05]0.956——————
Age-2.79[-4.73, -0.86]0.005\mathbf{0.005}-2.72[-4.43, -1.02]0.002\mathbf{0.002}-2.95[-4.61, -1.29]0.001\mathbf{0.001}
Short FES-I———-1.24[-5.61, 3.13]0.577———
BTrackS——————0.14[-0.93, 1.22]0.794
Table 7: SW Variability Analysis
 Model 1Model 2Model 3
PredictorsEstimatesCIpEstimatesCIpEstimatesCIp
(Intercept)393.99[110.09, 677.89]0.007\mathbf{0.007}594.80[330.99, 858.61]<0.001\mathbf{<0.001}602.10[330.27, 873.93]<0.001\mathbf{<0.001}
STEADI-19.62[-30.44, -8.80]<0.001\mathbf{<0.001}——————
Age-0.59[-4.61, 3.43]0.772-2.50[-6.27, 1.27]0.193-3.44[-7.21, 0.34]0.074
Short FES-I———-12.76[-22.43, -3.09]0.010\mathbf{0.010}———
BTrackS Balance Score——————-1.97[-4.42, 0.47]0.113
Table 8: SL Variability Analysis
 Model 1Model 2Model 3
PredictorsEstimatesCIpEstimatesCIpEstimatesCIp
(Intercept)26.82[5.86, 47.79]0.012\mathbf{0.012}13.39[-7.16, 33.95]0.20013.69[-7.46, 34.83]0.203
STEADI1.23[0.45, 2.01]0.002\mathbf{0.002}——————
Age0.21[-0.08, 0.51]0.1570.36[0.07, 0.65]0.016\mathbf{0.016}0.44[0.15, 0.74]0.003\mathbf{0.003}
Short FES-I———0.71[-0.01, 1.42]0.052———
BTrackS Balance Score——————0.00[-0.19, 0.19]0.989
Table 9: Sit-to-Stand and Stand-to-Sit (STS) Duration Analysis
More in this topic

Gait Analysis

Gait analysis uses movement and biomechanical signals to characterize walking, recognize people, assess health, and guide rehabilitation or locomotion.

See all 75 papers

Explore similar work

CardsList
Feb 24, 2026

Automating Timed Up and Go Phase Segmentation and Gait Analysis via the tugturn Markerless 3D Pipeline

Instrumented Timed Up and Go (TUG) analysis can support clinical and research decision-making, but robust and reproducible markerless pipelines are still limited. We present tugturn.py, a Python-based workflow for 3D markerless TUG processing that combines phase segmentation, gait-event detection, spatiotemporal metrics, intersegmental coordination, and dynamic stability analysis. The pipeline uses spatial thresholds to segment each trial into stand, first gait, turning, second gait, and sit phases, and applies a relative-distance strategy to detect heel-strike and toe-off events within valid gait windows. In addition to conventional kinematics, tugturn provides Vector Coding outputs and Extrapolated Center of Mass (XCoM)-based metrics. The software is configured through TOML files and produces reproducible artifacts, including HTML reports, CSV tables, and quality-assurance visual outputs. A complete runnable example is provided with test data and command-line instructions. This manuscript describes the implementation, outputs, and reproducibility workflow of tugturn as a focused software contribution for markerless biomechanical TUG analysis.

Abel Gonçalves Chinaglia, Guilherme Manna Cesar, Paulo Roberto Pereira Santiago

Biomechanics and Motor Control Laboratory, School of Physical Education and Sport of Ribeirao Preto, University of Sao Paulo, Brazil · Graduate Program in Rehabilitation and Functional Performance, Ribeirao Preto Medical School, University of Sao Paulo, Brazil · Department of Physical Therapy, Brooks College of Health, University of North Florida, USA

Gait AnalysisMarkerless Motion Capture
AskPDF
Mar 31, 2026

Sit-to-Stand Transitions Detection and Duration Measurement Using Smart Lacelock Sensor

Postural stability during movement is fundamental to independent living, fall prevention, and overall health, particularly among older adults who experience age-related declines in balance, muscle strength, and mobility. Among daily functional activities, the Sit-to-Stand (SiSt) transition is a critical indicator of lower-limb strength, musculoskeletal health, and fall risk, making it an essential parameter for assessing functional capacity and monitoring physical decline in aging populations. This study presents a methodology SiSt transition detection and duration measurement using the Smart Lacelock sensor, a lightweight, shoe-mounted device that integrates a load cell, accelerometer, and gyroscope for motion analysis. The methodology was evaluated in 16 older adults (age: mean: 76.84, SD: 3.45 years) performing SiSt tasks within the Short Physical Performance Battery (SPPB) protocol. Features extracted from multimodal signals were used to train and evaluate four machine learning classifiers using a 4-fold participant-independent cross-validation to classify SiSt transitions and measure their duration. The bagged tree classifier achieved an accuracy of 0.98 and an F1 score of 0.8 in classifying SiSt transition. The mean absolute error in duration measurement of the correctly classified transitions was 0.047, and the SD was 0.07 seconds. These findings highlight the potential of the Smart Lacelock sensor for real-world fall-risk assessment and mobility monitoring in older adults.

Md Rafi Islam, Md Rejwanul Haque, Elizabeth Choma +4

Department of Electrical and Computer Engineering, The University of Alabama, Tuscaloosa, AL 35487 · Genesis Medtech USA Inc · Whitworth University's Doctor of Physical Therapy program +2

Human Activity RecognitionWearable Health MonitoringWearable Sensing
AskPDF
Jul 16, 2026

Assessing Physical Frailty and Fall-Risk Indicators with Social Robots: An in situ Evaluation with Older Adults

Frailty assessments are crucial to evaluate the risk of adverse events and the health and social care needs of older adults, yet their administration remains resource-intensive and typically relies on coarse clinical outcomes, such as task completion times, which may overlook biomechanical indicators of functional decline. To address this, we present a robotic framework that guides older adults through standardised frailty and fall-risk tests while capturing clinical scores and additional frailty-related metrics, offering a deeper insight into a user's condition. The system uses a Behaviour Tree architecture that coordinates perception, decision-making, interaction, and measurement modules. Using vision-based skeleton tracking, the robot evaluates established clinical tests, including the Short Physical Performance Battery (SPPB) and the Timed Up and Go (TUG). The framework was co-designed with healthcare professionals and evaluated in situ during six months in a rehabilitation centre's research lab with N=81 older adults. Robot-derived measurements were compared against therapist assessments and clinical reference instruments, including a gait analysis walkway and an inertial measurement unit (IMU). Results showed excellent agreement for most test completion times and gait-related parameters (ICC>0.9ICC > 0.9). And, substantial agreement for the overall SPPB score comparing the robot and the therapist (k=0.67k = 0.67) and moderate agreement comparing the robot and the IMU (k=0.55k=0.55). The findings highlight that social robots can provide reliable and objective frailty assessments in healthcare settings while enabling the collection of relevant mobility indicators beyond conventional outcomes.

Aniol Civit, Antonio Andriella, Alba Martínez +4

Institut de Robòtica i Informàtica Industrial, CSIC-UPC, Llorens i Artigas 4-6, 08028 Barcelona, Spain · Parc Sanitari Pere Virgili, Area of Intermediate Care, Barcelona, Spain · RE-FiT Barcelona Research Group, Vall d’Hebron Institute of Research and Parc Sanitari Pere Virgili, Barcelona, Spain +1

Assistive RoboticsHuman-Robot InteractionClinical Risk Prediction
AskPDF