Key points
- Wearable inertial sensors are strongest for continuous, real-world data. In one study, people with Parkinson's disease and healthy older adults walked more slowly and more variably at home than in the lab.
- Force plates are the reference standard for ground reaction forces and balance, but they are fixed, expensive and confined to the settings that can house them.
- Markerless smartphone capture needs no hardware on the patient. Its spatiotemporal gait measures generally agree well with marker-based systems; some joint-angle and out-of-plane measures need more caution.
Three ways to measure the same movement
A clinician who wants objective movement data has more options than ever. Inertial sensors worn on the body, force plates built into the floor, and markerless capture that runs on a phone camera all produce numbers where there used to be an impression. They are not interchangeable, though. Each was designed around a different tradeoff between accuracy, setup, and where the measurement can happen—and understanding those tradeoffs matters more than chasing whichever tool sounds most advanced.
Smartphone vs. wearables vs. force plates at a glance
| Smartphone camera (markerless AI) | Wearable sensors (IMUs) | Force plates | |
|---|---|---|---|
| How it works | Computer vision estimates body landmarks from ordinary video | Accelerometers and gyroscopes worn on the body record motion signals | Sensors in a platform record ground reaction forces |
| What it measures best | Whole-body, visible tasks in a visit: gait, balance, sit-to-stand, range of motion | Continuous, real-world activity and walking over hours or days | Forces and balance (center of pressure) |
| What the patient wears | Nothing | One or more sensors, placed correctly | Nothing, but they must stand or move on the plate |
| Where it works | Exam room, bedside or home, wherever a phone can film | Anywhere the sensors are worn | Wherever the plate is installed |
| Main limitation | Some joint-angle and out-of-plane measures need more caution | Placement, charging, syncing and wear compliance | Fixed, expensive, confined to settings that can house them |
The sections below explain each row and the evidence behind it.
What wearables and force plates do well
Wearable inertial sensors are strong at continuous, in-the-wild data: step counts, activity over a day, cadence during a long walk. Strapped to the right landmark, they capture motion the clinic never sees — and that motion can differ from clinic performance. In one study using a single lower-back sensor, both people with Parkinson's disease and healthy older adults walked more slowly and more variably at home than in the lab (Del Din et al., 2016). Force plates, at the other end, are the reference standard for ground reaction forces and balance—precise, validated, and hard to beat for the specific things they measure.
The cost is friction. Wearables have to be placed correctly, charged, synced, and often worn for days, and placement drift changes the signal. Force plates are fixed, expensive, and confined to the settings that can house them. Both can produce excellent data, and both add steps that, on a full clinic day, quietly get skipped.
Where markerless smartphone capture fits
Markerless capture takes a different path: a standard phone camera, computer vision inferring movement across a set of body landmarks, and no hardware to attach to the patient. (We explain the technique in what markerless motion capture is.) Nothing to place, charge, or sync. That lowers the barrier to actually taking the measurement—which is often the real bottleneck, since the most accurate instrument is worthless for the assessment that never gets done.
The relevant question isn't which method wins in the abstract. It's whether a tool captures the movements a given assessment depends on, does it consistently, and fits the visit. (For how markerless capture actually compares to laboratory reference systems, see the peer-reviewed accuracy evidence.) In short: markerless spatiotemporal gait measures generally agree well with marker-based capture, while some joint-angle and out-of-plane measures need more caution (Wade et al., 2022). Smartphone-based gait analysis more broadly shows moderate-to-strong agreement with reference systems for parameters like gait speed and stride length, with results varying by condition, protocol, and device placement (Bea et al., 2025). For visual, whole-body tasks—gait, balance, sit-to-stand, range of motion—a camera sees the whole picture at once, without instrumenting the body.
What about consumer smartwatches and fitness trackers?
Many comparisons of "AI wearables vs. smartphone health tracking" are about consumer devices: smartwatches, rings and phone step counters. Those are built for general wellness tracking such as steps, activity and heart rate, worn all day. Clinical movement assessment asks a narrower question: how a specific patient performs a standardized test, such as gait speed over 10 meters or five timed chair stands, measured the same way at every visit so the numbers can be compared over time.
Either kind of device can use AI. A camera-based tool uses it to find body landmarks in video; a wearable uses it to interpret sensor signals. What separates a clinical measurement from a consumer reading is not the AI. It is whether the output has been validated against reference systems for that task, in that population, under a defined protocol.
Consistency and access, not just accuracy
A lab-grade number from an instrument that only comes out occasionally tells you less over time than a slightly simpler number captured the same way every visit. Trends are what clinicians act on, and trends require repeated, standardized measurement. Removing setup friction and letting assessment happen in a normal exam room—or at home—is what makes that repetition realistic. Read alongside clinical judgment, the best tool is the one that reliably produces comparable data at the moment care is being delivered, not the one with the most impressive spec sheet. What separates a clinical measurement from a consumer reading is validation and repeatability; see what makes movement measurement clinical-grade.
For the main tool types and named options in each, see our buyer's guide to movement analysis apps for clinicians.
References
- Del Din S, Godfrey A, Galna B, Lord S, Rochester L. Free-living gait characteristics in ageing and Parkinson's disease: impact of environment and ambulatory bout length. Journal of NeuroEngineering and Rehabilitation. 2016;13:46. doi:10.1186/s12984-016-0154-5
- Wade L, Needham L, McGuigan P, Bilzon J. Applications and limitations of current markerless motion capture methods for clinical gait biomechanics. PeerJ. 2022;10:e12995. doi:10.7717/peerj.12995
- Bea T, Chaabene H, Freitag CW, Schega L. Psychometric characteristics of smartphone-based gait analyses in chronic health conditions: a systematic review. Journal of Functional Morphology and Kinesiology. 2025;10(2):133. doi:10.3390/jfmk10020133