Interactive Systems · Motion Input
Phi Motion
An experimental camera-to-game input system that turns body-pose observations into discrete motion intents.
01 · Overview
Phi Motion combines a pose-provider boundary, neutral calibration, normalized motion features, intent recognition, and a Unity input bridge. The Runner integration translates a subset of intents into lane changes, jump, and slide.
02 · Product context
Noisy 2D pose frames do not directly become useful game controls. The system must establish a person-relative reference, interpret motion over time, and suppress stale or repeated commands.
03 · Intended users
Players and engineers exploring camera-based game interaction.
04 · My responsibility
Provider and native-to-Unity boundaries, calibration and intent logic, deterministic diagnostics, and Runner input integration.
05 · Constraints and accuracy
Apple Vision and MediaPipe providers are implemented, but the Runner's experimental default is MediaPipe with Demo Assist. Seven intent types exist; the Runner maps only left, right, jump, and crouch. Human physical acceptance and release reliability remain open.
06 · Verified technology and scope
Unity · iOS · Apple Vision · MediaPipe · C# · Native bridge
System evolution
How the system evolved
Pose observations
Native iOS and provider adapters deliver bounded 2D landmarks into a provider-neutral Unity pose representation.
Person-relative intent
Neutral calibration, normalized geometry, and temporal evidence separate deliberate movement from frame noise. Legacy and V3 recognizers remain experimental.
Game input
The Runner bridge accepts only eligible intent phases, bounds pending input, and maps four motions to existing game actions. Demo Assist is the current experimental default, not a release verdict.
Only verified information is included. No performance metrics, production capabilities, or outcomes have been inferred.