ESP32 + MPU-6050 motion sensor for velocity-based training data collection.

• ESP32 development board (BLE capable)
• GY-521 breakout (MPU-6050, 6-DoF)
• I2C connections: SDA=GPIO21, SCL=GPIO22

# Initialize ESP-IDF environment . ~/esp/esp-idf/export.sh # Build cd firmware idf.py build # Flash idf.py -p /dev/ttyUSB0 flash # Monitor serial output idf.py -p /dev/ttyUSB0 monitor

cd host pip install -r requirements.txt # Run (scans for device automatically) python barbell_buddy.py # With options python barbell_buddy.py --rate 100 --calibrate --show-samples # VBT mode for real-time velocity tracking python barbell_buddy.py --vbt --exercise squat --load-kg 100 # Help python barbell_buddy.py --help

Basic options:

• -r, --rate HZ: Stream rate 50-200 Hz (default: 100)
• -o, --output DIR: Output directory for logs (default: ./data)
• -c, --calibrate: Run gyro calibration before streaming
• -s, --show-samples: Show live acceleration values
• -d, --device ADDR: Skip scanning, connect to specific address

VBT (Velocity-Based Training) options:

• --vbt: Enable real-time velocity tracking and rep detection
• --exercise {squat,bench,deadlift}: Exercise type (metadata)
• --load-kg KG: Barbell load in kg (shown in output)
• --loss-threshold PCT: Warn when velocity loss exceeds threshold
• --invert: Invert velocity sign (if sensor mounted upside down)
• --no-bell: Disable terminal bell on rep completion

VBT tuning parameters:

• --tau SEC: Gravity filter time constant (default: 0.5)
• --stationary-accel-g G: Max accel deviation from 1g for rest detection (exercise-specific defaults)
• --stationary-gyro-dps DPS: Max gyro for rest detection (exercise-specific defaults)
• --stationary-hold-ms MS: Time to confirm stationary state (exercise-specific defaults)
• --move-confirm-ms MS: Time to confirm movement started (default: 50)
• --concentric-v-min M/S: Min velocity for concentric phase (default: 0.05)

Note: The --exercise flag automatically applies tuned thresholds for each lift. Bench press uses relaxed thresholds to handle the horizontal bar path and continuous movement pattern.

000001bb-0000-0010-8000-00805f9b34fb

• Stream (notify): 010001bb-0000-0010-8000-00805f9b34fb
• Command (write): 020001bb-0000-0010-8000-00805f9b34fb
• Status (read): 030001bb-0000-0010-8000-00805f9b34fb

Raw IMU data (imu_data_<timestamp>.csv):

timestamp_us,ax_mg,ay_mg,az_mg,gx_dps16,gy_dps16,gz_dps16,seq,flags 

Per-rep metrics (reps_<timestamp>.csv, VBT mode only):

rep_index,rep_duration_s,concentric_duration_s,peak_velocity_m_s,mean_velocity_m_s,peak_accel_m_s2,peak_gyro_dps,velocity_loss_pct 

Metadata JSON includes recording timestamp, firmware version, sensor ranges, exercise type, and load.

When --vbt is enabled, the tool computes real-time velocity-based training metrics:

• Live dashboard: Current velocity, rep count, stationary/moving state
• Rep detection: Automatically segments reps based on rest periods
• Per-rep metrics:
  • Mean Concentric Velocity (MCV): Time-weighted average during upward motion
  • Peak Concentric Velocity (PCV): Maximum velocity in concentric phase
  • Velocity Loss %: Fatigue indicator relative to best rep in set

Example output:

Rep 1 | MCV 0.52 m/s | PCV 0.68 m/s | Con 0.65s | Loss 0% Rep 2 | MCV 0.48 m/s | PCV 0.63 m/s | Con 0.71s | Loss 8% Rep 3 | MCV 0.41 m/s | PCV 0.55 m/s | Con 0.82s | Loss 21% WARNING: Velocity loss 21% exceeds threshold 20% 

The tool uses exercise-specific thresholds for optimal rep detection:

For best results with bench press:

• Pause briefly at lockout (~0.5s) between reps to ensure clean rep segmentation
• Touch-and-go style may merge consecutive reps together
• The algorithm detects rest periods to separate reps; without pauses, it sees continuous movement

For all exercises:

• Keep the sensor securely attached to the barbell
• Calibrate (-c) before each session for best accuracy
• Place the barbell at rest before starting your set