Can you measure gait with a smartphone camera?
Yes, for spatiotemporal parameters. Peer-reviewed work has derived gait measures from ordinary two-dimensional video and from smartphone video using pose estimation, reporting close agreement with laboratory reference systems for measures such as walking speed and step length. Accuracy depends on camera placement, frame rate, spatial calibration, and an unobstructed view of the walking path.
Last reviewed .
What the research shows
Work published in PLOS Computational Biology demonstrated two-dimensional video-based analysis of human gait using pose estimation, deriving gait parameters from ordinary single-camera video rather than from a marker-based laboratory setup.
The OpenCap work, also in PLOS Computational Biology, estimated human movement dynamics from smartphone videos, using consumer devices as the capture hardware.
A study examining the effect of walking speed on the reliability of a smartphone-based markerless gait analysis system reported excellent agreement with laboratory motion capture for spatiotemporal variables across speeds, with lower reliability for discrete joint range-of-motion measures, particularly at the hip and ankle at higher speeds.
What a phone camera can and cannot do well
| Measure | Support in published work |
|---|---|
| Walking speed, cadence, step and stride length, step and stride time | Consistently the strongest area of agreement with reference systems. |
| Stance and swing timing | Generally good agreement, though measures depending on narrow lateral distances agree less well. |
| Sagittal plane hip and knee kinematics | Moderate to good agreement reported across several studies. |
| Ankle kinematics, frontal and transverse plane angles | Consistently the weakest, and not reliably recovered from a single view. |
| Absolute joint angle quantification | Multi-camera three-dimensional setups provide higher precision than single-camera approaches. |
What a usable recording requires
The camera being ordinary does not mean the recording conditions can be. Most of the error in video-based gait measurement comes from the setup rather than from the model.
- A spatial reference — without a known real-world distance in the scene, image measurements cannot be converted into meters.
- A consistent camera position and angle, held the same between sessions that will be compared.
- Adequate frame rate, since gait events are short and are located in time by frame.
- An unobstructed view of the whole walking path, with the person fully in frame.
- Reasonable, even lighting and a background the person is distinguishable from.
Published accuracy figures are obtained under controlled recording conditions. A phone in an uncontrolled setting is not the same measurement instrument as a phone in a standardized protocol.
Where CurveAssure fits
CurveAssure records through a browser-based portal on a device a clinic or study site already owns, which is what makes the spatial reference necessary — the calibration floor marker supplies the known real-world distance the algorithm needs. Recordings upload for processing and are not left on the device.
CurveCapture is for in-clinic use and is prescription only. At-home capture is available for research use through CurveResearch where a study design allows; at-home assessment for clinical use is not available yet.
References
- Stenum J, Rossi C, Roemmich RT. Two-dimensional video-based analysis of human gait using pose estimation. PLOS Computational Biology, 2021. https://doi.org/10.1371/journal.pcbi.1008935
- Uhlrich SD, Falisse A, Kidziński Ł, et al.. OpenCap: human movement dynamics from smartphone videos. PLOS Computational Biology, 2023. https://doi.org/10.1371/journal.pcbi.1011462
- Accuracy, validity, and reliability of markerless camera-based 3D motion capture systems versus marker-based systems in gait analysis: a systematic review and meta-analysis. Sensors, 2024. https://doi.org/10.3390/s24113686
- Validity of AI-driven markerless motion capture for spatiotemporal gait analysis in stroke survivors. Sensors, 2025. https://doi.org/10.3390/s25175315