Sensor Data Fusion with Spresense and CommonSense

void setup() {
    pinMode(LED0, OUTPUT);
    pinMode(LED1, OUTPUT);
    pinMode(LED2, OUTPUT);
    pinMode(LED3, OUTPUT);
}
void loop() {
    digitalWrite(LED0, HIGH);
    delay(100);
    digitalWrite(LED1, HIGH);
    delay(100);
    digitalWrite(LED2, HIGH);
    delay(100);
    digitalWrite(LED3, HIGH);
    delay(1000);
    digitalWrite(LED0, LOW);
    delay(100);
    digitalWrite(LED1, LOW);
    delay(100);
    digitalWrite(LED2, LOW);
    delay(100);
    digitalWrite(LED3, LOW);
    delay(1000);
}

edge-impulse-daemon –-clear

#define FREQUENCY_HZ 0.1

#include <Arduino.h>
#include <SDHCI.h>
#include <File.h>
#include <LPS22HHSensor.h>
#include <HTS221Sensor.h>
#include <SensirionI2CSgp40.h>
#include <Artekit_APDS9250.h>
#include <Wire.h>
// Definitions for SD Card
SDClass SD;
File myFile;
// Definitions for LPS22HH
#define dev_interface       Wire
LPS22HHSensor PressTemp(&dev_interface);
// Definitions for HTS221
HTS221Sensor  HumTemp(&dev_interface);
// Definitions for SGP40
SensirionI2CSgp40 sgp40;
// Definitions for APDS9250
Artekit_APDS9250 myApds9250;
#define FREQUENCY_HZ        0.1
#define INTERVAL_MS         (1000 / (FREQUENCY_HZ))
static unsigned long last_interval_ms = 0;
static unsigned long sample_time = 0;
void setup()
{
  Serial.begin(115200);
  while (!Serial) {}

  // Initialize LED
  pinMode(LED_BUILTIN, OUTPUT);

  // Initialize I2C bus.
  dev_interface.begin();
  Wire.begin();


  // Initialize HTS221
  HumTemp.begin();
  HumTemp.Enable();

  // Initialize LPS22HH
  PressTemp.begin();
  PressTemp.Enable();

  // Initialize SGP40
  sgp40.begin(Wire);


  // Initialize APS9250
  myApds9250.begin();
  myApds9250.setMode(modeColorSensor);
  myApds9250.setResolution(res18bit);
  myApds9250.setGain(gain1);
  myApds9250.setMeasurementRate(rate100ms);

   // Initialize SD
  Serial.print("Insert SD card.");
  if (!SD.begin(4)) {
    Serial.println("SD Error");
    return;
  }
  Serial.println("SD Started");
  if(!SD.exists("datalog.csv"))
  {
      myFile = SD.open("datalog.csv", FILE_WRITE);
      if (myFile) {
        myFile.println("count,pres,temp,humi,voc,red,green,blue,ir");
        myFile.close();
      } else {
        Serial.println("Error creating datalog.csv");
      }
  }
}

int cnt = 0;


void loop()
{
  if (millis() > last_interval_ms + INTERVAL_MS) {
    digitalWrite(LED_BUILTIN, HIGH);
    last_interval_ms = millis();
    myFile = SD.open("datalog.csv", FILE_WRITE);

    if (myFile) {
      logData();
      cnt = cnt+1;
      digitalWrite(LED_BUILTIN, LOW);
    } else {
      Serial.println("Error opening file");
    }
  }
}


void logData(){

  float pressure, temp;
  PressTemp.GetPressure(&pressure);
  PressTemp.GetTemperature(&temp);

  float humidity;
  HumTemp.GetHumidity(&humidity);

  uint16_t defaultRh = 0x8000;
  uint16_t defaultT = 0x6666;
  uint16_t srawVoc = 0;
  uint16_t error = sgp40.measureRawSignal(defaultRh, defaultT, srawVoc);

  uint32_t red, green, blue, ir;
  myApds9250.getAll(&red, &green, &blue, &ir);

  Serial.print("Writing to datalog.csv");
  myFile.print(cnt);
  myFile.print(",");
  myFile.print(pressure);
  myFile.print(",");
  myFile.print(temp);
  myFile.print(",");
  myFile.print(humidity);
  myFile.print(",");
  myFile.print(srawVoc);
  myFile.print(",");
  myFile.print(red);
  myFile.print(",");
  myFile.print(green);
  myFile.print(",");
  myFile.print(blue);
  myFile.print(",");
  myFile.println(ir);
  myFile.close();

  /* For tests only

  Serial.print(" ==> Count=");
  Serial.print(cnt);
  Serial.print(", pressure=");
  Serial.print(pressure);
  Serial.print(", temperature=");
  Serial.print(temp);
  Serial.print(", humidity=");
  Serial.print(humidity);
  Serial.print(", voc=");
  Serial.print(srawVoc);
  Serial.print(", red =");
  Serial.print(red);
  Serial.print(", green =");
  Serial.print(green);
  Serial.print(", blue =");
  Serial.print(blue);
  Serial.print(", ir =");
  Serial.println(ir);
  */
}

def normalize_train_data(df):
    """
    Normalizes the numerical features of a dataset and returns a tuple
    containing three elements:
    1. The normalized data
    2. A list of the mins of each column.
    3. A list of the ranges of each column.
    """
    mins=[]
    ranges=[]
    # get numerical features
    df_scaled = df.iloc[:, :-1]

    # apply normalization
    for column in df_scaled.columns:
        min = df_scaled[column].min()
        range = (df_scaled[column].max() - min)
        df_scaled[column] = (df_scaled[column] - min) / range

        # Collect min and range values
        mins.append(min)
        ranges.append(range)

    # Combine the normalized features and output into a new data frame
    df_scaled['class'] = df['class']

    return df_scaled, mins, ranges

def normalize_test_data(df, mins, ranges):
    # Select the numerical columns to be standardized
    numerical_cols = df.columns[0:-1]

    # normalizeize the numerical columns
    df[numerical_cols] = (df[numerical_cols] - mins) / ranges

    return df

scaled_train_df.to_csv("scaled_train_df.csv", index=False)
scaled_test_df.to_csv("scaled_test_df.csv", index=False)

with open("df_mins.txt", "w") as f:
for s in df_mins:
f.write(str(s) +"\n")

with open("df_ranges.txt", "w") as f:
for s in df_ranges:
f.write(str(s) +"\n")

static const float features[] = {
0.8987, 0.1213, 0.3832, 0.8012, 0.5352, 0.5698, 0.6240, 0.4511, 0.9295, 0.1227, 0.3832, 0.7951, 0.5363, 0.5708, 0.6240, 0.4541, 0.9251, 0.1255, 0.3832, 0.7956, 0.5374, 0.5729, 0.6240, 0.4557
};

#include <Arduino.h>
#include <LPS22HHSensor.h>
#include <HTS221Sensor.h>
#include <SensirionI2CSgp40.h>
#include <Artekit_APDS9250.h>
#include <Wire.h>


// Edge Impulse Library
#include <CommonSense-Sensor-Fusion-Preprocessed-data-v2_inferencing.h>

// Definitions for LPS22HH
#define dev_interface       Wire
LPS22HHSensor PressTemp(&dev_interface);
// Definitions for HTS221
HTS221Sensor  HumTemp(&dev_interface);
// Definitions for SGP40
SensirionI2CSgp40 sgp40;
// Definitions for APDS9250
Artekit_APDS9250 myApds9250;


// Settings
#define DEBUG               1           // 1 to print out debugging info
#define DEBUG_NN            false       // Print out EI debugging info
#define ANOMALY_THRESHOLD   0.3         // Scores above this are an "anomaly"
#define SAMPLING_FREQ_HZ    1             // Inference sampling frequency (Hz)
#define SAMPLING_PERIOD_MS  1000 / SAMPLING_FREQ_HZ // Inf. samp. period (ms)
#define NUM_SAMPLES         EI_CLASSIFIER_RAW_SAMPLE_COUNT // 3 sample @ 0.1 Hz
#define READINGS_PER_SAMPLE EI_CLASSIFIER_RAW_SAMPLES_PER_FRAME // 8


// above definitions come from model_metadata.h

// Preprocessing constants
float mins[] = {
  895.85, 21.02, 30.2, 30066, 5, 4, 1, 4
};
float ranges[] = {
  2.27, 7.10, 10.7, 2133, 954, 1000, 511, 659
};



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

  // Initialize I2C bus.
  dev_interface.begin();
  Wire.begin();
  // Initialize HTS221
  HumTemp.begin();
  HumTemp.Enable();
  // Initialize LPS22HH
  PressTemp.begin();
  PressTemp.Enable();
  // Initialize SGP40
  sgp40.begin(Wire);
  // Initialize APS9250
  myApds9250.begin();
  myApds9250.setMode(modeColorSensor);
  myApds9250.setResolution(res18bit);
  myApds9250.setGain(gain1);
  myApds9250.setMeasurementRate(rate100ms);
}



void loop() {

  float pressure, temp;
  float humidity;
  uint16_t defaultRh = 0x8000;
  uint16_t defaultT = 0x6666;
  uint16_t srawVoc = 0;
  uint32_t red, green, blue, ir;
  unsigned long timestamp;
  static float raw_buf[NUM_SAMPLES * READINGS_PER_SAMPLE];
  static signal_t signal;
  int max_idx = 0;
  float max_val = 0.0;
  char str_buf[40];

  // Collect samples and perform inference
  for (int i = 0; i < NUM_SAMPLES; i++) {
    // Take timestamp so we can hit our target frequency
    timestamp = millis();

    // read from sensors
    PressTemp.GetPressure(&pressure);
    PressTemp.GetTemperature(&temp);
    HumTemp.GetHumidity(&humidity);
    uint16_t error = sgp40.measureRawSignal(defaultRh, defaultT, srawVoc);
    myApds9250.getAll(&red, &green, &blue, &ir);

    // Store raw data into the buffer
    raw_buf[(i * READINGS_PER_SAMPLE) + 0] = pressure;
    raw_buf[(i * READINGS_PER_SAMPLE) + 1] = temp;
    raw_buf[(i * READINGS_PER_SAMPLE) + 2] = humidity;
    raw_buf[(i * READINGS_PER_SAMPLE) + 3] = srawVoc;
    raw_buf[(i * READINGS_PER_SAMPLE) + 4] = red;
    raw_buf[(i * READINGS_PER_SAMPLE) + 5] = green;
    raw_buf[(i * READINGS_PER_SAMPLE) + 6] = blue;
    raw_buf[(i * READINGS_PER_SAMPLE) + 7] = ir;

    // Perform preprocessing step (normalization) on all readings in the sample
    for (int j = 0; j < READINGS_PER_SAMPLE; j++) {
      temp = raw_buf[(i * READINGS_PER_SAMPLE) + j] - mins[j];
      raw_buf[(i * READINGS_PER_SAMPLE) + j] = temp / ranges[j];
    }

    // Wait just long enough for our sampling period
    while (millis() < timestamp + SAMPLING_PERIOD_MS);
  }

   // Print out our preprocessed, raw buffer
#if DEBUG
  for (int i = 0; i < NUM_SAMPLES * READINGS_PER_SAMPLE; i++) {
    Serial.print(raw_buf[i]);
    if (i < (NUM_SAMPLES * READINGS_PER_SAMPLE) - 1) {
      Serial.print(", ");
    }
  }
  Serial.println();
#endif

  // Turn the raw buffer in a signal which we can the classify
  int err = numpy::signal_from_buffer(raw_buf, EI_CLASSIFIER_DSP_INPUT_FRAME_SIZE, &signal);

  if (err != 0) {
      ei_printf("ERROR: Failed to create signal from buffer (%d)\r\n", err);
      return;
  }
  // Run inference
  ei_impulse_result_t result = {0};
  err = run_classifier(&signal, &result, DEBUG_NN);
  if (err != EI_IMPULSE_OK) {
      ei_printf("ERROR: Failed to run classifier (%d)\r\n", err);
      return;
  }

  // Print the predictions
  ei_printf("Predictions ");
  ei_printf("(DSP: %d ms., Classification: %d ms., Anomaly: %d ms.)\r\n",
        result.timing.dsp, result.timing.classification, result.timing.anomaly);

  for (int i = 0; i < EI_CLASSIFIER_LABEL_COUNT; i++) {
    ei_printf("\t%s: %.3f\r\n",
              result.classification[i].label,
              result.classification[i].value);
  }

  // Print anomaly detection score
#if EI_CLASSIFIER_HAS_ANOMALY == 1
  ei_printf("\tanomaly acore: %.3f\r\n", result.anomaly);
#endif

  // Find maximum prediction
  for (int i = 0; i < EI_CLASSIFIER_LABEL_COUNT; i++) {
    if (result.classification[i].value > max_val) {
      max_val = result.classification[i].value;
      max_idx = i;
    }
  }
}