A swipe can be done with the forearm alone; being short does not make it recognizable

What it is

Biomechanically a swipe can live below the elbow: a forearm throw, a little wrist, almost no shoulder. That is short, cheap, and easier to hide socially. Short is not the same as well recognized. If the sensor was modelled on full-arm waves in amplitude and duration, a forearm swipe falls into noise or under the segmentation window. Copy that says “just a light swipe” while the recognizer waits for a larger path produces field behaviour that works sometimes. Short is a feasible lower bound on the action, not an automatic passport for recognition.

Why it happens

The observables of a ballistic swipe are peak speed, duration, and spatial span. Forearm-only spans are often a dozen centimetres and can last under 150 ms. Distant skeletons already jitter at centimetre scale, so short paths have poor signal-to-noise. At 30 Hz the whole token is four or five frames and the direction histogram is unstable. Near-field hand tracking can see the stretch, but if thresholds still follow the large waves in the demo, a short swipe is labelled preparation or jitter. The more practised people are, the shorter they dare to go; if templates came from first-timers' large moves, experts fail more. Stability is a sensing-resolution and template-scale problem, not a “did they swipe” problem.

Studying it

Have the same person produce the same direction at three scales: wrist only, forearm, whole arm. Report recognition, direction errors, and misses. Log peak speed, span, and frame count. Train templates on large moves versus mixed scales and watch recall on short swipes. Measure distant cameras and near-field headsets separately. Add a social measure: did a bystander notice the move? Hiding is often why short swipes are designed; that dimension does not replace the recognition number.

Where it stops holding

Near-field headsets with millimetre fingertip tracking can make forearm or even finger flicks stable; living-room skeleton tracking cannot. Adding a pinch edge (pinch then throw) stabilizes immediately, but that is no longer a pure swipe. Children's whole-arm span equals an adult forearm; adult whole-arm templates will treat a child's whole arm as “too short.” After fatigue, shoulders refuse to move and practised users spontaneously switch to a forearm strategy. A system that only accepts large moves goes blind in the second half of the task.

Applying it

• Teaching should state a minimum acceptable span (for example “the hand travels more than a palm width”), not only “a light swipe.”
• Templates must include forearm-scale samples. If a distant device cannot see the forearm, do not promise small moves in the copy.
• Accept binned by wrist / forearm / whole arm. If only whole-arm passes, either raise sensing or publish the minimum amplitude.

Related

• Same group: C4.07.1 A swipe is a throw-out plus a return · C4.07.3 Teaching must keep direction, sequence, and reset · C4.07.4 Whether firing waits for the return is an implementation choice
• Adjacent: C4.04 Static poses versus dynamic gestures · C4.12 Fatigue cost of mid-air gestures
• Search: short swipe · forearm flick · gesture amplitude threshold