ESP32 Arduino library for connecting to a Myo armband via Bluetooth Low Energy (BLE). Enables real-time streaming of EMG (electromyography) and IMU sensor data.

• BLE connection to Myo armband with auto-reconnect
• Auto-scan by BLE name — no MAC address needed (myo myo;)
• EMG data streaming (8 channels, 2 samples per notification packet)
• IMU data parsing — Quaternion, Accelerometer, Gyroscope (int16 raw + scale factors)
• Euler angles — Roll / Pitch / Yaw in degrees computed from quaternion
• Pose / Gesture detection — Rest, Fist, WaveIn, WaveOut, FingersSpread, DoubleTap
• Arm sync state — detects which arm (left/right) and sync status
• Unlock / Lock control — prevents auto-lock during data collection
• Vibration control (short / medium / long)
• Battery level monitoring
• CSV output for AI/ML training — labeled samples via Serial command
• Python data collector (collect_data.py) — saves labeled CSV to disk automatically

• ESP32 development board
• Myo armband (Thalmic Labs)

Install via Arduino IDE Board Manager / Library Manager:

• ESP32 Arduino Core
• BLEDevice, BLEClient, BLEUtils, BLEScan (included in ESP32 core)

#include "myo.h"

// Option A: auto-scan (finds any Myo by BLE name – no address needed)
myo myo;

// Option B: fixed address (faster reconnect if you know the MAC)
// myo myo("f9:64:aa:5e:d8:ef");

void setup() {
  Serial.begin(115200);
}

void loop() {
  myo.connect();           // auto-reconnect if disconnected
  myo.getAllData();        // start EMG + IMU + classifier streaming
  myo.unlock();          // hold-unlock: prevents auto-lock
  myo.EMGNotify();
  myo.IMUNotify();
  myo.BATTNotify();
  myo.PoseNotify();

  // Serial commands: 0-9 = record labeled CSV sample, h = header, s/m/l = vibrate
  if(Serial.available()) {
    char cmd = Serial.read();
    if(cmd >= '0' && cmd <= '9') myo.printCSVLine(cmd - '0');
    else if(cmd == 'h')          myo.printCSVHeader();
    else if(cmd == 's')          myo.vibrate(1);
    else if(cmd == 'm')          myo.vibrate(2);
    else if(cmd == 'l')          myo.vibrate(3);
  }
  delay(10);
}

Send the CSV header once, then type 0–9 over Serial Monitor to record labeled samples:

h          → label,e0s0,...,e7s1,qw,qx,qy,qz,ax,ay,az,gx,gy,gz,pose
0          → 0,12,-3,0,8,5,-1,22,-7,9,2,1,6,3,4,18,-4,16100,...,0
1          → 1,45,30,...

Copy the output into a .csv file for training with scikit-learn, TensorFlow, etc.

float roll, pitch, yaw;
myo.getEuler(roll, pitch, yaw);
Serial.printf("Roll: %.1f  Pitch: %.1f  Yaw: %.1f\n", roll, pitch, yaw);

int8_t   emg[2][8];    // 2 samples x 8 channels
int16_t  imu_quat[4];  // w, x, y, z
int16_t  imu_accel[3]; // x, y, z
int16_t  imu_gyro[3];  // x, y, z
MyoPose  currentPose;  // REST, FIST, WAVE_IN, WAVE_OUT, FINGERS_SPREAD, DOUBLE_TAP
bool     armSynced;    // true after arm-sync gesture
uint8_t  arm;          // 0 = right, 1 = left

collect_data.py receives labeled samples from the ESP32 and saves them to a CSV file.

pip install pyserial
python collect_data.py

Then in the Arduino Serial Monitor:

• h — send CSV header (once at the start)
• 0–9 — record a sample with that label
• s / m / l — vibrate for tactile feedback during recording

The script shows a live counter per label and prints ESP32 status messages (Pose:, Battery:, etc.) separately. Press Ctrl+C to stop and see the label distribution.

• BLE auto-scan by device name — myo myo; finds any Myo without knowing the MAC address
• Unlock/lock commands — unlock() sends hold-unlock to prevent auto-lock during sessions
• Arm sync state — parsePose() now handles all classifier events (sync, unsync, lock, unlock)
• Euler angles — getEuler(roll, pitch, yaw) converts quaternion to degrees
• Python data collector — collect_data.py saves labeled CSV samples automatically

• IMU data fully parsed: Quaternion, Accelerometer, Gyroscope (int16 raw, documented scale factors)
• Pose / gesture detection via classifier events (Fist, WaveIn/Out, FingersSpread, DoubleTap, Rest)
• Vibration command: vibrate(1/2/3) for short / medium / long
• Auto-reconnect on BLE disconnect via BLEClientCallbacks
• CSV output for AI/ML training data collection (printCSVHeader() / printCSVLine(label))
• Fixed getAllData() mode command: IMU mode was 0 (off), corrected to 3 (send all)
• All 4 EMG characteristics now registered with callbacks
• Serial command interface for vibration and CSV recording

• Fixed buffer overflow in fw_serial_number (serial number is 6 bytes, not 7)
• Fixed fw_reserved array size (8 entries)
• Fixed off-by-one in EMG raw-byte debug loop (< instead of <=)
• Added null-checks for all BLE service/characteristic lookups
• connect() now checks return value and only sets connected = true on success
• BLE device address is now configurable via constructor (no recompile needed)
• EMG notification callback implemented with correct 8-channel signed-int parsing
• Corrected getMyoInfo() field byte offsets to match Myo protocol