SIMCAP turns motion into meaning

A project for gesture-recognition using inertial measurement. Sensor Inferred Motion CAPture.

Hardware: 9-DoF IMU streaming at 50 Hz over Bluetooth Low Energy, running on espruino.

Software pipeline:

  • Real-time sensor visualization
  • Labeled gesture collection for training data
  • Template-based recognition via 3D spatial matching
  • ML pipeline with tflite export for on-device inference

Multiple firmware variants (GAMBIT, MOUSE, KEYBOARD, BAE) enable different input modalities—gestures become keystrokes, mouse movements, or custom triggers.

The insight: small motions carry semantic weight when you build the vocabulary.

Live: simcap.parc.land

See also: dygram, parc-land