Arduino IDE is the program you use to write and upload code to your ESP32.

Download

1. Go to: arduino.cc/en/software
2. Click "Download" for your operating system
3. Install it like any other app

For Mac Users

1. Download the .dmg file
2. Drag Arduino to Applications
3. First time: Right-click > Open (to bypass Gatekeeper)

For Windows Users

1. Download the .exe installer
2. Run the installer
3. Check "Install USB driver" when prompted

Arduino doesn't know about ESP32 by default. We need to add it.

Add ESP32 Board URL

1. Open Arduino IDE
2. Mac: Arduino IDE > Settings (or ⌘ + ,)
3. Windows: File > Preferences
4. Find "Additional boards manager URLs"
5. Paste this URL:

6. Click OK

Install ESP32 Board

1. Go to Tools > Board > Boards Manager
2. Search for "esp32"
3. Find "ESP32 by Espressif Systems"
4. Click Install (this takes 2-5 minutes)

Select Your Board

1. Go to Tools > Board > ESP32 Arduino
2. Select "ESP32 Dev Module"

Now we upload the code that makes your ESP32 read the AD8232 sensors.

Select Your Port

1. Plug in your ESP32 via USB
2. Go to Tools > Port
3. Select the port that appears (usually has "USB" or "UART" in the name)

The Code

Delete everything in the editor and paste this:

// 3-high-speed-capture.ino
// N-Channel Adaptive EMG Capture (250Hz+)
// Add or remove pins to match your hardware.

// ==========================================
// CHANNEL CONFIGURATION
// ==========================================
const int CHANNEL_PINS[] = {34, 36};  // GPIO pins for each AD8232
// Examples:
//   1 channel (chin only):       {34}
//   2 channels (chin + throat):  {34, 36}
//   3 channels (chin + L + R):   {34, 36, 35}
const int NUM_CHANNELS = sizeof(CHANNEL_PINS) / sizeof(CHANNEL_PINS[0]);

const unsigned long SAMPLE_INTERVAL_MICROS = 4000; // 250 Hz

unsigned long lastSampleTime = 0;

void setup() {
    Serial.begin(115200);
    for (int i = 0; i < NUM_CHANNELS; i++) {
        pinMode(CHANNEL_PINS[i], INPUT);
    }
    analogReadResolution(12);
    analogSetAttenuation(ADC_11db);
    Serial.println("READY");
}

void loop() {
    unsigned long currentMicros = micros();
    if (currentMicros - lastSampleTime >= SAMPLE_INTERVAL_MICROS) {
        lastSampleTime = currentMicros;

        // CSV Format: Timestamp, Ch1, Ch2, ..., ChN
        Serial.print(millis());
        for (int i = 0; i < NUM_CHANNELS; i++) {
            Serial.print(",");
            Serial.print(analogRead(CHANNEL_PINS[i]));
        }
        Serial.println();
    }
}

Upload

1. Click the → Upload button (top left arrow)
2. Wait for "Compiling..." then "Uploading..."
3. You'll see progress percentages
4. Done when you see "Done uploading"

Now let's see if your ESP32 is actually sending data.

Open Serial Monitor

1. In Arduino IDE, click the magnifying glass icon (top right)
2. Or go to Tools > Serial Monitor
3. Set baud rate to 115200 (dropdown at bottom)

What You Should See

Understanding the Numbers

Python lets us analyze the EMG data more powerfully than the Serial Monitor.

Check if Python is Installed

Open Terminal (Mac/Linux) or Command Prompt (Windows) and type:

If you see Python 3.x.x, you're good. If not, install it:

Install Python

1. Go to: python.org/downloads
2. Download the latest Python 3.x
3. Windows: Check "Add Python to PATH" during install!
4. Install

Install Required Packages

In Terminal/Command Prompt, run:

This Python script checks if all your channels are working correctly — without electrodes.

Create the Script

1. Create a new file called validation_test.py
2. Paste the code below
3. Change the port to match yours (from Arduino IDE Tools > Port)

Run It

1. CLOSE Arduino Serial Monitor
2. Open Terminal/Command Prompt
3. Navigate to where you saved the file
4. Run: python3 validation_test.py

Expected Output

Now we test with electrodes on your body to verify muscle signals are detected.

1. Skin Preparation (Crucial!)

Poor skin contact = noise. Do not skip this.

• Wipe: Vigorously wipe chin, throat, and earlobe with an alcohol pad.
• Dry: Wait 10 seconds for it to dry completely.
• Placement: Stick electrodes before connecting cables.

2. Electrode Placement Strategy

We target the speech-relevant musculature identified in the AlterEgo paper (Kapur et al., 2018). Precision here determines signal quality.

1. Channel 1: Mental Region (Chin)

Primary Signal. Captures Mentalis and Digastric muscles.

2. Channel 2: Left Laryngeal (Throat) (Kapur Table 1, Rank #2)

Swallow—feel your Adam's apple move UP? That's the laryngeal region. Place electrodes on the LEFT side.

3. Channel 3: Right Laryngeal (Throat) (Kapur Table 1, Rank #3)

Mirror of Channel 2. Right side of Adam's apple.

3. Cable Connections

• Yellow + Green: Snap onto the muscle site (e.g., Chin). Polarity doesn't matter for differential EMG.
• Red (RL): Connect ONE red snap to your Earlobe. Leave the other two red snaps hanging (disconnected).

Create the Test Script

Save this as electrode_test.py:

Run It

What the Tests Check

Now that your hardware is verified, it's time to build your dataset. We need to record examples of you "speaking" specific words silently.

Usage

1. This step is simpler now!
2. Go to the ML Walkthrough (Tab 4).
3. We will use the advanced 4-curriculum-recorder.py there.
4. It handles everything: labels, phases, and saving.

Congratulations! Your AlterEgo hardware and software are working. Here's what comes next:

Immediate Next Steps

1. Data Collection: Record samples of you "silently speaking" different words
2. Feature Extraction: Convert raw signals to MFCC features
3. Model Training: Teach AI to recognize your words
4. Real-time Inference: Use the trained model to decode silent speech

Expected Performance

Resources

• Hardware Guide: alterego-step.html
• Original Paper: Kapur et al., "AlterEgo" (2018)
• ESP32 Docs: espressif.com/esp32

Ensure you have the valid High-Speed CSV Firmware running (from Software Step 3).

Run the Python script. It will ask you for session details.

cd code/python
python3 4-curriculum-recorder.py

Session Setup

1. Subject ID: Enter your name (e.g., carl).
2. Phase: Select 3 (Mouthing). This is the best starting point.
3. Labels: Enter: ONE, TWO, THREE, SILENCE (comma-separated).
4. Samples: Enter 20 (20 reps per word — research minimum).

What You'll See

The live values update in real-time. Watch them jump when you flex! Channel count adapts to your hardware.