Wearable
Integration
Apple Watch, Wear OS, HarmonyOS — three independent watch apps with on-device AI shot recognition and professional sports science metrics.
On-Device AI Shot Recognition
CoreML / TFLite on-device inference, 6 shot types classified in real-time, latency < 300ms
Live Heart Rate
Precise real-time heart rate tracking with automatic Zone 0-5 classification
Movement Heatmap
On-court movement visualization for positioning efficiency and stamina analysis
Shot Recognition
1D-CNN model (100 frames × 6 channels), real-time classification of 6 shot types
Serve Speed Estimation
AI-estimated serve speed (±15% margin), tracking your serve improvement curve
Energy Expenditure
Scientific calorie calculation with heart rate zone-based workload assessment
Auto Data Sync
Automatic post-session sync to phone, offline cache for up to 30 sessions
Supported Device Matrix
Tier 1 Full Version
On-device AI shot recognition, fully offline capable
Tier 2 Lite Version
Phone-based inference, basic data collection
Professional Sports Science Metrics
Based on professional athlete training monitoring methodology for quantifiable training quality assessment
TRIMP (Training Impulse)
Per SessionBanister HR formula to quantify per-session training load
CTL (Chronic Training Load)
42-day exponential weighted average reflecting long-term fitness
ATL (Acute Training Load)
7-day exponential weighted average reflecting recent training stress
ACWR (Acute:Chronic Ratio)
CoreATL/CTL ratio, ideal range 0.8-1.3, alerts for injury risk
Training Pyramid
Low-intensity aerobic 70-80% / High-intensity 15-20% / Anaerobic 5-10%
Effort Score
Per-session composite load score (0-10), AI provides training quality assessment
Comprehensive Health Monitoring
Resting Heart Rate
7/30/90-day trend charts
HRV
RMSSD reflecting autonomic balance
VO2max
iOS HealthKit aerobic capacity
Recovery Score
0-100 score for daily intensity
Heart Rate Recovery
1-min post-exercise HR drop
HR Zone Analysis
Zone 0-5 duration & percentage
Fully Offline Capable
Tier 1 devices run on-device AI models directly on the watch — no phone or network needed.