The Problem With Manual Rep Counting
Every lifter has been there. Set 3 of heavy squats. You are grinding. Was that rep 7 or 8? You cannot remember because your body is focused on not collapsing. So you either do an extra rep to be safe or end the set early.
Automatic rep counting solves this. Your Apple Watch counts for you, accurately and consistently, regardless of how hard you are working.
The Hardware: What Your Apple Watch Contains
Accelerometer — Measures acceleration forces in three dimensions (x, y, z). Detects the upward acceleration of your wrist during a curl, the pause at the top, and the controlled deceleration down.
Gyroscope — Measures rotational velocity around three axes. Captures angular movement of your wrist, helping distinguish between different movement patterns.
Combined, these sensors generate data at 50-100 Hz — millions of data points during a 45-minute workout.
The Software Layer: CoreMotion
Apple provides CoreMotion for processed sensor data. REPVEX uses:
- CMMotionManager — raw accelerometer and gyroscope data in real-time
- CMDeviceMotion — fused output separating user acceleration from gravity
The Signal Processing Pipeline
Step 1: Noise Filtering
A low-pass filter removes high-frequency noise while preserving the fundamental movement pattern.Step 2: Feature Extraction
Peak detection, amplitude analysis, period measurement, and waveform shape analysis.Step 3: Pattern Matching
Each exercise has a signature movement pattern. The app uses exercise-specific models to match detected patterns.Step 4: Rep Validation
Minimum amplitude threshold, minimum time between reps, and consistency checks prevent false counts.Accuracy Across Exercises
| Exercise | Accuracy | Why |
|---|---|---|
| Bicep curls | 95%+ | Large, consistent wrist movement |
| Bench press | 93%+ | Clear up-down pattern |
| Overhead press | 93%+ | Distinct vertical acceleration |
| Rows | 90%+ | Pull-toward pattern detectable |
| Squats | 88%+ | Wrist moves less but pattern is consistent |
| Lateral raises | 92%+ | Strong rotational signal |
| Deadlifts | 85%+ | Minimal wrist-specific movement |
The Role of Machine Learning
Modern implementations like REPVEX use ML models trained on thousands of labeled workout sessions. The model can classify exercises, predict rep boundaries even in noisy signals, and adapt to individual lifting styles. Combined with heart rate zone tracking, the Apple Watch becomes a comprehensive training computer.
All processing happens locally on the Apple Watch — zero latency, no internet required, battery efficient.
Try It Yourself
Download REPVEX free, pair your Apple Watch, and do a set of bicep curls. Watch the rep counter tick up in real-time on your wrist.