All research

Wearable biomechanics

Estimating 3D foot kinetics from wearable signals

How much of the force and moment profile measured by an instrumented treadmill can be recovered from insole signals?

IEEE Journal of Biomedical and Health Informatics · 2026 (online 2025)

  1. Smart-insole pressure signals
  2. Estimated forces and moments
  3. Comparison with treadmill measurements

Approach

Ke2KeNet uses time series from 16 insole pressure sensors per foot. The study trained separate models for three force components, three moment components, and two centre-of-pressure coordinates, then compared the estimates with treadmill recordings across walking and running speeds.

Reported results

For the average gait cycles in Table I, the reported Pearson correlation exceeded 0.95 for all target signals except the right-foot vertical moment.

Healthy participants in cross-validation
9
Subject-independent folds
5
Estimated signals per foot
8

My contribution

I led preprocessing, model development, subject- and speed-wise evaluation, analysis, and manuscript preparation.

Study scope

The results come from nine healthy participants under controlled treadmill conditions. The plot illustrates one participant’s walking trials and shows where the estimated moment departs from the reference.

Ongoing work

I’m investigating how to combine plantar sensing with inertial measurements from the foot and shank, and how well models transfer across datasets.

Mediolateral ground reaction moment

Normalized amplitude (unitless)

The estimated moment is above the treadmill mean for most of the right-foot walking cycle, with the largest separation around the middle of the cycle.

Gait cycle (%)

Participant 10, right foot, walking at 0.9–5.4 km/h. The black line is the treadmill mean and the red line is the Ke2KeNet estimate. Shading shows one standard deviation on either side of each mean. This example comes from the archived analysis for Figure 5b and shows a visible mid-cycle discrepancy. Archived analysis notebook

Run the insole workflow

Use Python 3.12 and the tagged code release v0.1.0. The demo uses a small U-Net and synthetic signals to exercise the insole workflow on CPU without study data or pretrained weights.

  1. Get the release

    git clone --branch v0.1.0 --single-branch https://github.com/Sakib1263/Unimodal-3D-Kinetics-Estimation-from-Insole-or-IMU.git
    cd Unimodal-3D-Kinetics-Estimation-from-Insole-or-IMU
  2. Install and run the demo

    Create an environment with python -m venv .venv. Activate it with source .venv/bin/activate on Linux or macOS, or .\.venv\Scripts\Activate.ps1 in PowerShell. Then run the following commands.

    python -m pip install -r requirements.txt
    python -m kinetics demo --config configs/insole.json
  3. Check the output

    The command prints Saved run: …/runs/demo_Fy_LL_UNet_…. Open that directory and check that status.json reports "status": "complete". The run also writes summary.json and fold-level predictions. Those results describe synthetic inputs.

Work with the study data

The processed treadmill and insole recordings are publicly downloadable. Use the data preparation guide to create the expected input files. Then inspect the target, scale, and participant splits before training.

python -m kinetics inspect --config configs/insole.json

The release contains the maintained workflow. The article’s training-ready exports and pretrained checkpoints are separate from the release.