NTZ Conveyor Belt Proof of Concept
Research collaboration with Petr Janata, UC Davis
What We're Building
A tool to analyze where skiers look while navigating mogul fields.
By combining eye-tracking data with body motion sensors, we can understand how
expert skiers plan their routes through complex terrain.
The Core Concept
Phase Angle
Each ski turn follows a compression/extension cycle synced to the mogul terrain.
We represent this as a continuous 0-360° phase angle:
- 0° — Maximum compression at mogul crest (turn pivot)
- 180° — Maximum extension in trough
- 360° — Next crest, turn complete
Next Turn Zones (NTZ)
The skier's visual field is divided into zones representing where future turns might occur:
- NTZ-0 — Current turn location
- NTZ-1 — Next turn (one ahead)
- NTZ-2 — Two turns ahead
- NTZ-3+ — Further look-ahead
The Conveyor Belt Model
As the skier progresses through each turn, the NTZ bands "scroll" toward them —
like the Star Wars opening crawl. When a turn completes (phase crosses 360°),
the zones decrement: NTZ-1 becomes NTZ-0, NTZ-2 becomes NTZ-1, etc.
┌─────────────────────────────────────┐
│ NTZ-3 (far ahead) │ ↓
├─────────────────────────────────────┤ ↓
│ NTZ-2 │ ↓ scrolling
├─────────────────────────────────────┤ ↓ toward
│ NTZ-1 (next turn) │ ↓ skier
├─────────────────────────────────────┤ ↓
│ NTZ-0 (current turn) │ ↓
└─────────────────────────────────────┘
▲
SKIER
Research Questions
- How far ahead do expert skiers look? (NTZ-1? NTZ-2? NTZ-3?)
- Does gaze pattern change with turn phase? (look ahead mid-turn, focus near at pivot?)
- Do experts and novices show different look-ahead patterns?
- Can we identify optimal gaze strategies for training?
Data Pipeline
Pupil Labs Glasses IMU Sensor Drone Camera
│ │ │
▼ ▼ │
Gaze (x,y) Accel/Gyro/Quaternion │
│ │ │
│ ▼ ▼
│ Phase Angle Mogul Field Map
│ (0-360°) (absolute)
│ │ │
└────────────┬───────────────┘ │
▼ │
Frame-by-Frame Data │
(gaze + phase + turn#) │
│ │
▼ │
NTZ Classification ◄───────────────────────────┘
(which zone was gaze in?)
│
▼
Aggregated Analysis
(patterns across turns/runs/skiers)
Current Status
- Data model: Complete (Mission → Day → Session → Run → Segment → Turn → Frame → GazePoint)
- API: Working, returns frame-level and summary data
- Visualization: Prototype with synthetic data
- Real data: Awaiting sample from Petr's lab
Open Questions
Several design decisions depend on clarification from the research team.
See InitialPlanWithQuestions.md
for the full list.
Links