Capture architecture · the Apple sensors you already wear

The gait lab you already wear.

A phone in your pocket. A watch on your wrist. AirPods playing a song. That's three motion sensors at three points on your body — a distributed gait lab you're already wearing, right now, while you walk to the kitchen. You don't need a new device. You need the software that reads all three at once and turns them into one measured gait signal.

Three capture points on the body A figure showing AirPods at the head, iPhone at the pocket, and Apple Watch at the wrist as three motion-sensor locations. AirPodsthe head iPhonethe pelvis Apple Watchthe wrist
01 · The proof it's already real

Your AirPods are already a shipping body sensor.

Branch Ergonomics, a work-wellness app on the App Store today, reads the motion sensor inside your AirPods to catch a slouch in real time and buzz you to sit up. It uses CMHeadphoneMotionManager — the exact head-tracking IMU Apple built for spatial audio. The lesson is the whole thesis: the head is now an instrumented body segment on a device hundreds of millions of people already wear. If that sensor is good enough to detect posture, it is good enough to measure how your head moves while you walk. The sensor shipped. The gait application is open.

02 · Three points on one body

Head, pelvis, wrist — each tells a different truth about your gait.

Single-point sensing gives you a number. Three points at different segments of the body give you coordination — the thing a clinician actually reads when they watch someone walk.

POINT 1

AirPods — the headnovel

The head sits atop the kinetic chain and houses the vestibular system — the body's own balance sensor. How smoothly the head travels through a stride is a direct read on gait stability.

CMHeadphoneMotionManager · ~25 Hz
Signal: head-acceleration smoothness & regularity (harmonic ratio, step-to-step consistency) — established markers of stability and fall risk. Almost nobody measures this in everyday walking.
POINT 2

iPhone — the pelvis

In a pocket, the phone rides near the body's center of mass — the workhorse location of the entire phone-gait literature. This is where speed and rhythm live.

CMMotionManager · CMPedometer
Signal: cadence, step length, double-support, vertical oscillation, speed — computed from raw motion, from first principles, not read from a black box.
POINT 3

Apple Watch — the wrist

The wrist captures arm swing — and reduced or asymmetric arm swing is one of the earliest, most sensitive markers of a gait or neurological change.

CMMotionManager (watchOS)
Signal: arm-swing amplitude & symmetry, independent cadence — a second witness that confirms the pelvis reading and adds the limb the phone can't see.
03 · Why three beats one

The signal isn't in any one sensor. It's in how they agree.

One IMU tells you how fast. Three IMUs at head, pelvis, and wrist tell you how well — because gait quality is a coordination problem, and coordination only shows up across body segments.

What fusion unlocks

Head-vs-pelvis decoupling — a stable walker's head stays quiet while the pelvis sways; when the head starts tracking every wobble, stability is failing. Arm-swing vs trunk symmetry — a limp or a one-sided change shows as a mismatch. Three independent cadences that agree — confidence you can't get from one clock. This is the difference between a pedometer and a gait assessment.

And you own the computation

Apple hands you a pre-cooked walkingSpeed scalar. We ran the method on real Apple Health data and watched that scalar arrive mislabeled — imperial values tagged as metric. Raw tri-IMU means you compute speed, symmetry, and smoothness yourself, validate them against the published method, and never inherit a black box's unit bug. End-to-end signal you control.

04 · The honest engineering

What's easy today, what's a fast-follow, what's the limit.

Phone + AirPods = today

Both streams surface through the phone's Core Motion at once — a synchronized two-point capture (pelvis + head) is buildable now, no watch required. The phone is the master clock; AirPods samples arrive phone-timestamped.

The watch is a fast-follow

The wrist adds arm swing via a small watchOS companion that streams to the phone and aligns to under ~100 ms — ample when footfalls land roughly once a second. A third point, not a blocker to shipping the first two.

A guided walk, not 24/7

AirPods head-motion runs while the AirPods are in and the app is active (~25 Hz; it pauses for calls). So this is a 60-second guided walk — "pocket the phone, keep your AirPods in, walk the hall and back" — not silent background sensing. A weekly measured walk is the clinical cadence anyway.

Consumer-grade, stated plainly

Three phones-and-earbuds IMUs are not an instrumented gait lab, and this is decision support, not a diagnosis — the same honest posture as the method. It is tuned for trend and coordination, where consumer sensing is genuinely strong.

05 · Where it sits — one engine, richer input

A capture tier, not a new engine.

New sensors don't change the method — they enrich what it reads. The 3-point capture is the differentiated middle tier between the passive phone read and clinical hardware.

TIER 0
Live today
The HealthKit read

Passive, zero effort, pre-cooked Apple Health mobility metrics. The on-ramp — running now at healthgait.com/check.

TIER 1
This page
The 3-point capture

A guided walk, raw multi-IMU from AirPods + iPhone + Watch, computed from first principles. The layer you already wear — and the one nobody else is building.

TIER 2
Clinical hardware

Plantar-pressure insoles, powered assist — from partners, when the clinical bar rises. Same method, deeper signal.

06 · For device, shoe & hearable partners

The capture layer that turns worn devices into a measured signal.

You make the hardware people wear. We make it mean something.

Earbuds, a smart watch, a shoe with a sensor — each is a stream of raw motion that, alone, doesn't screen, doesn't bill, and doesn't carry a defensible claim. This capture architecture fuses worn devices into one gait signal and runs it through a published method — a number, a risk band, a decline flag, a reimbursable event. Bring your device, or ride the three your customer already owns.

Start a partner conversation → Read the method
About this page. This describes a capture architecture for informational purposes and is not medical advice, not a diagnosis, and not a substitute for professional clinical evaluation. The MotionSole measurement software, as software alone, is not an FDA-cleared medical device; any diagnostic or fall-prevention claim is subject to separate regulatory clearance. Sensor APIs, sampling rates, and device behaviors described (CMHeadphoneMotionManager, CMMotionManager, CMPedometer) reflect general public developer information as understood at drafting and change across OS versions — verify against current Apple documentation before building. Apple, AirPods, Apple Watch, iPhone, and HealthKit are trademarks of Apple Inc.; MotionSole is not affiliated with, endorsed by, or sponsored by Apple Inc. Branch Ergonomics is named only as a publicly available example of AirPods motion sensing and is not affiliated with MotionSole. Consumer-grade motion sensing is less precise than instrumented gait analysis; this method is tuned for trend and coordination.