WristWrist Trainer

Under the hood

No camera, no wearable - so what does the counting?

This guide breaks Wrist Trainer's counting into three stages: sensing, calibration, and decision. Everything runs on the phone itself - and knowing why helps explain when a rep might be missed.

A three-stage pipeline

From raw signals to one clean count.

01

Sense: two inner ears

The accelerometer reads gravity direction and linear acceleration; the gyroscope reads rotation speed. Together they track how the phone's orientation changes.

02

Calibrate: memorize your level line

Held sideways with the screen down for three seconds, the app records gravity at that instant as neutral. Every angle afterwards is an estimate relative to it.

03

Decide: pass the target, come back

After the relative angle passes the target and holds for the set time, the rep stays pending until the phone returns to the neutral zone. Jitter around the threshold never stacks extra reps.

Design trade-off

Prefer one count too few over one too many.

The neutral-return rule trades a little combo speed for the guarantee that every count matches one full back-and-forth movement. Per-second haptics show hold progress, a triple pulse confirms completion, and spoken cues announce each count.

  1. Off-axis feedback is immediate. If the pose drifts out of the starting plane, the app asks you to re-place the phone instead of counting blindly.
  2. Sensing and deciding happen in on-device memory. Raw sensor samples are never stored or uploaded.
  3. Only summaries are saved. Local history keeps per-direction totals, not continuous waveforms.
  4. Online features say so explicitly. Sign-in and purchase recovery use the network; training and counting never depend on either.