Mental Health Monitoring System

The system uses three main sensors:

1. MAX30102 - Optical heart rate and blood oxygen sensor
2. GSR Sensor - Galvanic Skin Response sensor for stress detection
3. BH1750 - Ambient light sensor

These sensors work together to collect physiological data that can indicate mental health status.

// Sensor Objects in the code
MAX30105 particleSensor;  // MAX30102 sensor object
BH1750 lightMeter;       // BH1750 light sensor
const int GSR_PIN = 34;  // GSR sensor pin

The sensors are connected as follows:

• MAX30102 and BH1750 use I2C communication (connected to SDA and SCL pins)
• GSR sensor is connected to analog pin 34

The code initializes these connections in the setup() function:

void setup() {
  Wire.begin();  // Initialize I2C
  particleSensor.begin(Wire, I2C_SPEED_FAST); // MAX30102
  lightMeter.begin(); // BH1750
  // GSR is analog so only needs pin declaration
}

The MAX30102 sensor detects heart rate by measuring changes in blood volume using optical technology (PPG). The code implements:

1. IR LED shines light into the skin
2. Photodetector measures reflected light
3. Pulse detection algorithm identifies beats
4. Beat-to-beat intervals are calculated for HRV

if(irValue > 50000) {  // Finger detected
  if(checkForBeat(irValue)) {
    long delta = millis() - lastBeat;
    lastBeat = millis();
    beatsPerMinute = 60000.0 / delta;
  }
}

Based on Shaffer et al. (2014) HRV measurement standards

The moving average calculation separates the GSR signal into two components:

• Tonic component (slow-changing baseline)
• Phasic component (fast-changing responses)

This is implemented with a 5-sample moving average window:

// Moving average for EDA tonic (Boucsein, 2012)
float edaMovingAverage(float newVal) {
  eda_buffer[buffer_index] = newVal;
  buffer_index = (buffer_index + 1) % WINDOW_SIZE;
  float sum = 0;
  for(int i=0; i<WINDOW_SIZE; i++) sum += eda_buffer[i];
  return sum/WINDOW_SIZE;
}

HRV is calculated using SDNN (Standard Deviation of NN intervals), which is a well-established metric in psychophysiology research (Castaldo et al., 2019).

The code implements this by:

1. Storing the last 16 RR intervals (time between heartbeats)
2. Calculating the mean interval
3. Computing the standard deviation of these intervals

// Calculate SDNN (Standard Deviation of NN intervals)
float calculateSDNN() {
  float mean = 0.0f;
  for (int i = 0; i < RR_BUFFER_SIZE; i++) {
    mean += rrIntervals[i];
  }
  mean /= RR_BUFFER_SIZE;
  // ... variance calculation ...
  return sqrt(variance);
}

HRV is an important indicator of autonomic nervous system activity and stress levels.

The system uses a neural network (defined in MentalHealthNN.h) that takes sensor inputs and outputs probabilities for four mental states:

1. Normal
2. Stress
3. Anxiety
4. Depression

The prediction happens in the detectMentalHealth() function:

void detectMentalHealth(float edaTonic, float edaPhasic, float bpm, float hrv, float lux, 
    MentalHealthStatus& status, float* probabilities) {
  // Get prediction from neural network
  status = mentalHealth.predict(edaTonic, edaPhasic, bpm, hrv, lux);
  // Get probabilities for each class
  mentalHealth.getPredictionProbabilities(edaTonic, edaPhasic, bpm, hrv, lux, probabilities);
}

The neural network was likely trained on labeled physiological data from individuals with known mental health conditions.

The BH1750 light sensor measures ambient light levels which are important because:

• Light exposure affects circadian rhythms and mood (Golden et al., 2005)
• Low light levels may correlate with depressive symptoms
• Sudden light changes can indicate environmental stressors

The light data is read and included in the mental health analysis:

// Read BH1750 ambient light sensor
float lux = lightMeter.readLightLevel();

// Included in neural network input
detectMentalHealth(..., lux, ...);

The system uploads sensor data to GitHub repositories every 10 seconds using the GitHub API:

1. Data is formatted as JSON documents
2. Base64 encoded for GitHub API requirements
3. Uploaded using HTTP PUT requests with authentication
4. SHA hashes are tracked for version control

// Example of GitHub upload function
void uploadToGitHub(DynamicJsonDocument doc, const char* url, String& lastSHA) {
  HTTPClient http;
  http.begin(url);
  http.addHeader("Content-Type", "application/json");
  http.addHeader("Authorization", token);

  String jsonStr;
  serializeJson(doc, jsonStr);
  String encodedData = base64::encode(jsonStr);

  String payload = "{\"message\":\"Update data\",\"content\":\"" + 
      encodedData + "\",\"sha\":\"" + lastSHA + "\"}";

  int httpCode = http.PUT(payload);
  // ... handle response ...
}

Four separate JSON files are maintained for each sensor type and the mental status.