Home

Minds-On: Digital

In Short

An iterative exploration of head-movement interaction (nodding & shaking) to control a tilting board and rolling ball. Starting from a physical prototype, I moved into Unity for digital prototyping, and later integrated TouchDesigner for smoother signals and exhibition-ready visuals.

# Natural Interaction# Embodied Interaction
Minds-On Digital overview

Physical Roots (Brief)

The project began as a tangible tilt-board with a rolling ball to validate the embodied interaction concept. For full details, see the dedicated page: Minds-On →

Physical Minds-On prototype

Unity Digital Prototype

I rebuilt the tilt-board in Unity to allow faster iteration and rule testing. Head nodding maps to forward/backward tilt, while shaking maps to left/right tilt. I implemented clamped angles and smoothing to keep motion stable and responsive.

1) MediaPipe Face landmark detection

The MediaPipe Face Landmarker task lets user detect face landmarks and facial expressions in images and videos. More info →

MediaPipe face landmark detection

2) MediaPipe, Natively in Unity

Using MediaPipe inside Unity (create task → open camera → feed frames → run inference → visualize)

MediaPipe running natively in Unity

3) Calculation: Nods & Shakes → Board Tilt

Direction: From landmarks — nose vs. vertical center gives forward/back (pitch sign); left ear vs. right ear z gives left/right (yaw sign).
Angle: Normalize and gain these offsets to get a [-1,1] intensity, and smooth to produce the final tilt.

Head movement calculation

Unity + TouchDesigner Hybrid

For richer visuals and smoother input, I integrated TouchDesigner. TD handles preprocessing and visualization; Unity focuses on physics and game logic. This modular pipeline is ideal for exhibitions and scalable to multi-user setups.

1) Workflow in TouchDesigner

Use a community-made MediaPipe component/plugin.

TouchDesigner workflow

2) Facial feature extraction

Read landmarks for nose tip, left/right eyes, ears; Compute Left/Right from the eyes’ vertical difference and Up/Down from the nose–ear vertical difference, normalize and smooth them.

Facial feature extraction

3) OSC

Stream the four directional intensities via OSC to Unity for tilt control.

OSC data sent to Unity

Comparison

Comparison of physical, Unity-only, and hybrid prototypes