esp32S3AI模块做方言识别控制用电器项目的历程

1.测试esp32S3AI模块麦克风功能
用下面代码实现捕获声音信号，板载麦克风引脚分配图

<font size="4">#include <ESP_I2S.h>
I2SClass I2S;

void setup() {

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

  // setup 38 PDM clock and 39 PDM data pins
  I2S.setPinsPdmRx(38, 39);

  // start I2S at 16 kHz with 16-bits per sample
  if (!I2S.begin(I2S_MODE_PDM_RX, 16000, I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_MONO)) {
    Serial.println("Failed to initialize I2S!");
    while (1); // do nothing
  }
}

void loop() {
  // read a sample
  int sample = I2S.read();

  if (sample && sample != -1 && sample != 1) {
    Serial.println(sample);
  }
}</font>
复制代码

板子配置如上图
烧录代码后，在串口绘图可观察到声音变化图形

表明代码与开发板麦克风工作正常。


2.实现将录制的声音保存到 microSD 卡

<font size="4">#include "ESP_I2S.h"
#include "FS.h"
#include "SD.h"
#include "SPI.h"

void setup() {
  I2SClass i2s;

  uint8_t *wav_buffer;
  size_t wav_size;

  Serial.begin(115200);
  while (!Serial) {
    delay(10);
  }

  Serial.println("Initializing I2S bus...");

  i2s.setPinsPdmRx(38,39);

  // start I2S at 16 kHz with 16-bits per sample
  if (!i2s.begin(I2S_MODE_PDM_RX, 16000, I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_MONO)) {
    Serial.println("Failed to initialize I2S!");
    while (1); // do nothing
  }

  Serial.println("I2S bus initialized.");
  Serial.println("Initializing SD card...");

#ifdef REASSIGN_PINS
  SPI.begin(sck, miso, mosi, cs);
  if (!SD.begin(cs)) {
#else
  if (!SD.begin()) {
#endif
    Serial.println("Card Mount Failed");
    return;
  }
  uint8_t cardType = SD.cardType();

  if (cardType == CARD_NONE) {
    Serial.println("No SD card attached");
    return;
  }
  Serial.println("SD card initialized.");
  Serial.println("Recording 20 seconds of audio data...");

  // Record 20 seconds of audio data
  wav_buffer = i2s.recordWAV(20, &wav_size);

  // Create a file on the SD card
  File file = SD.open("/Arduinor_rec.wav", FILE_WRITE);
  if (!file) {
    Serial.println("Failed to open file for writing!");
    return;
  }

  Serial.println("Writing audio data to file...");

  // Write the audio data to the file
  if (file.write(wav_buffer, wav_size) != wav_size) {
    Serial.println("Failed to write audio data to file!");
    return;
  }

  // Close the file
  file.close();

  Serial.println("Application complete.");
}

void loop() {
  delay(1000);
  Serial.printf(".");
}</font>
复制代码

打开串口监视器，当串口输出信息”Recording 20 seconds of audio data...“时，模块开始录音，当串口显示”Writing audio data to file...“录音结束。随后出现............时，断电取下SD卡，用电脑查看音频文件arduinor_rec.wav，播放该文件，音量较小。


采集声音到SD卡，可增加音量，代码如下：

<font size="4">#include "ESP_I2S.h"
#include "FS.h"
#include "SD.h"
#include "SPI.h"

// make changes as needed
#define RECORD_TIME   10  // seconds, The maximum value is 240
#define WAV_FILE_NAME "data"
#define I2S_READ_CHUNK_SIZE 1920  // 添加到文件头部或全局作用域‌:ml-citation{ref="1,6" data="citationList"}

// do not change for best
#define SAMPLE_RATE 16000U
#define SAMPLE_BITS 16
#define WAV_HEADER_SIZE 44
#define VOLUME_GAIN 4 //取4音量较大，3应该更合适

I2SClass I2S;
String baseFileName;

int fileNumber = 1;
bool isRecording = false;

void setup() {
  
  Serial.begin(115200);
  while (!Serial) ;
  
  // setup 38 PDM clock and 39 PDM data pins
  I2S.setPinsPdmRx(38,39);
  if (!I2S.begin(I2S_MODE_PDM_RX, 16000, I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_MONO)) {
    Serial.println("Failed to initialize I2S!");
    while (1) ;
  }

 #ifdef REASSIGN_PINS
  SPI.begin(sck, miso, mosi, cs);
  if (!SD.begin(cs)) {
#else
  if (!SD.begin()) {
#endif
    Serial.println("Card Mount Failed");
    return;
  }
  uint8_t cardType = SD.cardType();
  if (cardType == CARD_NONE) {
    Serial.println("No SD card attached");
    return;
  }
  Serial.println("SD card initialized.");

  Serial.printf("Enter with the label name\n");
  //record_wav();
}

void loop() {
  if (Serial.available() > 0) {
    String command = Serial.readStringUntil('\n');
    command.trim();
    if (command == "rec") {
      isRecording = true;
    } else {
      baseFileName = command;
      fileNumber = 1; // reset file number each time a new base file name is set
      Serial.printf("Send rec for starting recording label \n");
    }
  }
  if (isRecording && baseFileName != "") {
    String fileName = "/" + baseFileName + "." + String(fileNumber) + ".wav";
    fileNumber++;
    record_wav(fileName);
    delay(1000); // delay to avoid recording multiple files at once
    isRecording = false;
  }
}

void record_wav(String fileName)
{
  uint32_t sample_size = 0;
  uint32_t record_size = (SAMPLE_RATE * SAMPLE_BITS / 8) * RECORD_TIME;
  uint8_t *rec_buffer = NULL;
  Serial.printf("Start recording ...\n");
   
  File file = SD.open(fileName.c_str(), FILE_WRITE);
  // Write the header to the WAV file
  uint8_t wav_header[WAV_HEADER_SIZE];
  generate_wav_header(wav_header, record_size, SAMPLE_RATE);
  file.write(wav_header, WAV_HEADER_SIZE);

  // PSRAM malloc for recording
  rec_buffer = (uint8_t *)ps_malloc(record_size);
  if (rec_buffer == NULL) {
    Serial.printf("malloc failed!\n");
    while(1) ;
  }
  Serial.printf("Buffer: %d bytes\n", ESP.getPsramSize() - ESP.getFreePsram());

  // 开始录音
  sample_size = 0;
  sample_size = I2S.readBytes((char*)rec_buffer, record_size);

  if (sample_size == 0) {
    Serial.println("Record Failed!");
  } else {
    Serial.printf("Record %d bytes\n", sample_size);
  }
  // Increase volume
  for (uint32_t i = 0; i < sample_size; i += SAMPLE_BITS/8) {
    (*(uint16_t *)(rec_buffer+i)) <<= VOLUME_GAIN;
  }

  // Write data to the WAV file
  Serial.printf("Writing to the file ...\n");
  if (file.write(rec_buffer, record_size) != record_size)
    Serial.printf("Write file Failed!\n");

  free(rec_buffer);
  file.close();
  Serial.printf("Recording complete: \n");
  Serial.printf("Send rec for a new sample or enter a new label\n\n");
}

void generate_wav_header(uint8_t *wav_header, uint32_t wav_size, uint32_t sample_rate)
{
  // See this for reference: http://soundfile.sapp.org/doc/WaveFormat/
  uint32_t file_size = wav_size + WAV_HEADER_SIZE - 8;
  uint32_t byte_rate = SAMPLE_RATE * SAMPLE_BITS / 8;
  const uint8_t set_wav_header[] = {
    'R', 'I', 'F', 'F', // ChunkID
    file_size, file_size >> 8, file_size >> 16, file_size >> 24, // ChunkSize
    'W', 'A', 'V', 'E', // Format
    'f', 'm', 't', ' ', // Subchunk1ID
    0x10, 0x00, 0x00, 0x00, // Subchunk1Size (16 for PCM)
    0x01, 0x00, // AudioFormat (1 for PCM)
    0x01, 0x00, // NumChannels (1 channel)
    sample_rate, sample_rate >> 8, sample_rate >> 16, sample_rate >> 24, // SampleRate
    byte_rate, byte_rate >> 8, byte_rate >> 16, byte_rate >> 24, // ByteRate
    0x02, 0x00, // BlockAlign
    0x10, 0x00, // BitsPerSample (16 bits)
    'd', 'a', 't', 'a', // Subchunk2ID
    wav_size, wav_size >> 8, wav_size >> 16, wav_size >> 24, // Subchunk2Size
  };
  memcpy(wav_header, set_wav_header, sizeof(set_wav_header));
}</font>
复制代码

上传后，在串口输入“hello",该字串将是声音文件的名称前缀，根据串口信息提示，输入”rec",开始录音，如此循环。

1.准备声音数据集
利用昨天最后一个程序实现声音收集。首先收集用建瓯方言（而且是乡村口音）说“开灯”关键词，方法是先格式化SD卡，插入ESP32S3AI模块SD卡槽，上电。打开Arduino IDE串口，等串口提示：“Send rec for a new sample or enter a new label”时，输入关键词的标签。由于每次录音只有10秒，我就多次采集该关键词，输入“on”,表示采集“开灯”语音录音文件。
此时，串口提示：，
在串口中输入“rec”,开始采集该段录音。


对着esp32s3用方言重复说“开灯”

10秒后结束本次录音。

用上述方法分别录制10次噪音、10次“关灯”，10次“开灯”。

用”split sample"把时长为10秒的文件分割成1秒的数据，方法如下图。

通过电脑麦克风采集的声音文件时，可以先标注收集数据的标签，如下图是将要采集的数据标签为off（关灯）

当三类声音的样本量都有60个时，观察每个声音数据的波形，删除一些不合格的文件，如下图

开关灯声音的正常的波形应该是下面这样的。

我的数据集分三个分类：开灯、关灯、背景音，每类51个样本。

3.准备好数据集，就可以进行设置Impulse

首先，impulse处理的音频文件为1秒窗口的数据点，每 500 毫秒滑动一次该窗口。 必须用小于 1 秒的零样本填充。另外，我们将每个 1 秒的音频样本经过预处理并转换为图像（例如，13 x 49 x 1）。因此，我使用 MFCC，它从音频信号中提取特征，这对分析人类的声音效果很好。模型各类是Classifier，使用卷积神经网络进行图像分类构建分类模型。

点击下面的按键后保存参数，此时左侧多出了MFCC 和 Classifier两个条目
                                                   


点击MFCC,显示MFCC是把声音变图像的功能.

   

点击“generate feature"生成特征按键，可观察三个分类声音文件分布图

现在要设置训练模型的超参数。


我的数据量小，训练速度很快就完成。



4.部署模型，首先要先生成Arduino 库文件

点击 build,右侧显示生成模型的过程

解压下载的模型文件

     把该文件安装到Arduino libraries中，打开“esp32_microphone”文件，并做适合的修改，修改后的文件如下：

<font size="4">#include <fj_dialect_detector_inferencing.h>
#include "ESP_I2S.h"
#define SAMPLE_RATE 16000U
#define SAMPLE_BITS 16

/** Audio buffers, pointers and selectors */
typedef struct {
    int16_t *buffer;
    uint8_t buf_ready;
    uint32_t buf_count;
    uint32_t n_samples;
} inference_t;

static inference_t inference;
static const uint32_t sample_buffer_size = 2048;
static signed short sampleBuffer[sample_buffer_size];
static bool debug_nn = false; // Set this to true to see e.g. features generated from the raw signal
static bool record_status = true;
I2SClass I2S;
/**
 * @brief      Arduino setup function
 */
void setup()
{
    // put your setup code here, to run once:
    Serial.begin(115200);
    // comment out the below line to cancel the wait for USB connection (needed for native USB)
    while (!Serial);
    Serial.println("Edge Impulse Inferencing Demo");
    pinMode(3, OUTPUT); // Set the pin as output
    digitalWrite(3, LOW); //Turn off
    
    // setup 38 PDM clock and 39 PDM data pins
    I2S.setPinsPdmRx(38, 39);
    if (!I2S.begin(I2S_MODE_PDM_RX, 16000, I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_MONO)) {
      Serial.println("Failed to initialize I2S!");
    while (1) ;
    }

    // summary of inferencing settings (from model_metadata.h)
    ei_printf("Inferencing settings:\n");
    ei_printf("\tInterval: ");
    ei_printf_float((float)EI_CLASSIFIER_INTERVAL_MS);
    ei_printf(" ms.\n");
    ei_printf("\tFrame size: %d\n", EI_CLASSIFIER_DSP_INPUT_FRAME_SIZE);
    ei_printf("\tSample length: %d ms.\n", EI_CLASSIFIER_RAW_SAMPLE_COUNT / 16);
    ei_printf("\tNo. of classes: %d\n", sizeof(ei_classifier_inferencing_categories) / sizeof(ei_classifier_inferencing_categories[0]));

    ei_printf("\nStarting continious inference in 2 seconds...\n");
    ei_sleep(2000);

    if (microphone_inference_start(EI_CLASSIFIER_RAW_SAMPLE_COUNT) == false) {
        ei_printf("ERR: Could not allocate audio buffer (size %d), this could be due to the window length of your model\r\n", EI_CLASSIFIER_RAW_SAMPLE_COUNT);
        return;
    }

    ei_printf("Recording...\n");
}

/**
 * @brief      Arduino main function. Runs the inferencing loop.
 */
void loop()
{
    bool m = microphone_inference_record();
    if (!m) {
        ei_printf("ERR: Failed to record audio...\n");
        return;
    }

    signal_t signal;
    signal.total_length = EI_CLASSIFIER_RAW_SAMPLE_COUNT;
    signal.get_data = µphone_audio_signal_get_data;
    ei_impulse_result_t result = { 0 };

    EI_IMPULSE_ERROR r = run_classifier(&signal, &result, debug_nn);
    if (r != EI_IMPULSE_OK) {
        ei_printf("ERR: Failed to run classifier (%d)\n", r);
        return;
    }

    int pred_index = 0;     // Initialize pred_index
    float pred_value = 0;   // Initialize pred_value

    // print the predictions
    ei_printf("Predictions ");
    //ei_printf("(DSP: %d ms., Classification: %d ms., Anomaly: %d ms.)",
     //   result.timing.dsp, result.timing.classification, result.timing.anomaly);
   // ei_printf(": \n");
    for (size_t ix = 0; ix < EI_CLASSIFIER_LABEL_COUNT; ix++) {
        ei_printf("    %s: ", result.classification[ix].label);
        ei_printf_float(result.classification[ix].value);
        ei_printf("\n");

    if (result.classification[ix].value > pred_value){
           pred_index = ix;
           pred_value = result.classification[ix].value;
    }
  }
    ei_printf("%d: ",pred_index);
    ei_printf("\n");
    if (pred_index == 1){
      digitalWrite(3, LOW); //Turn on
    }
    else if (pred_index == 2){
      digitalWrite(3, HIGH); //Turn off
    }

#if EI_CLASSIFIER_HAS_ANOMALY == 1
    ei_printf("    anomaly score: ");
    ei_printf_float(result.anomaly);
    ei_printf("\n");
#endif
}

static void audio_inference_callback(uint32_t n_bytes)
{
    for(int i = 0; i < n_bytes>>1; i++) {
        inference.buffer[inference.buf_count++] = sampleBuffer[i];

        if(inference.buf_count >= inference.n_samples) {
          inference.buf_count = 0;
          inference.buf_ready = 1;
        }
    }
}

static void capture_samples(void* arg) {

  const int32_t i2s_bytes_to_read = (uint32_t)arg;
  size_t bytes_read = i2s_bytes_to_read;

  while (record_status) {

    /* read data at once from i2s */
    //i2s_read((i2s_port_t)1, (void*)sampleBuffer, i2s_bytes_to_read, &bytes_read, 100);
    
    //I2S.i2s_read(esp_i2s::I2S_NUM_0, (void*)sampleBuffer, i2s_bytes_to_read, &bytes_read, 100);
    //从 I2S 接口中读取 i2s_bytes_to_read 字节到 sampleBuffer 中，读取结果保存在 bytes_read 中。
    bytes_read = I2S.readBytes((char*)sampleBuffer, i2s_bytes_to_read);

    if (bytes_read <= 0) {
      ei_printf("Error in I2S read : %d", bytes_read);
    }
    else {
        if (bytes_read < i2s_bytes_to_read) {
        ei_printf("Partial I2S read");
        }

        // scale the data (otherwise the sound is too quiet)
        for (int x = 0; x < i2s_bytes_to_read/2; x++) {
            sampleBuffer[x] = (int16_t)(sampleBuffer[x]) * 8;
        }

        if (record_status) {
            audio_inference_callback(i2s_bytes_to_read);
        }
        else {
            break;
        }
    }
  }
  vTaskDelete(NULL);
}

/**
 * @brief      Init inferencing struct and setup/start PDM
 *
 * @param[in]  n_samples  The n samples
 *
 * @return     { description_of_the_return_value }
 */
static bool microphone_inference_start(uint32_t n_samples)
{
    inference.buffer = (int16_t *)malloc(n_samples * sizeof(int16_t));

    if(inference.buffer == NULL) {
        return false;
    }

    inference.buf_count  = 0;
    inference.n_samples  = n_samples;
    inference.buf_ready  = 0;

    // if (i2s_init(EI_CLASSIFIER_FREQUENCY)) {
    //     ei_printf("Failed to start I2S!");
    // }

    ei_sleep(100);

    record_status = true;

    xTaskCreate(capture_samples, "CaptureSamples", 1024 * 32, (void*)sample_buffer_size, 10, NULL);

    return true;
}

/**
 * @brief      Wait on new data
 *
 * @return     True when finished
 */
static bool microphone_inference_record(void)
{
    bool ret = true;

    while (inference.buf_ready == 0) {
        delay(10);
    }

    inference.buf_ready = 0;
    return ret;
}

/**
 * Get raw audio signal data
 */
static int microphone_audio_signal_get_data(size_t offset, size_t length, float *out_ptr)
{
    numpy::int16_to_float(&inference.buffer[offset], out_ptr, length);

    return 0;
}

/**
 * @brief      Stop PDM and release buffers
 */
static void microphone_inference_end(void)
{
    free(sampleBuffer);
    ei_free(inference.buffer);
}


// static int i2s_init(uint32_t sampling_rate) {
//   // Start listening for audio: MONO @ 8/16KHz
//   i2s_config_t i2s_config = {
//       .mode = (i2s_mode_t)(I2S_MODE_MASTER | I2S_MODE_RX | I2S_MODE_TX),
//       .sample_rate = sampling_rate,
//       .bits_per_sample = (i2s_bits_per_sample_t)16,
//       .channel_format = I2S_CHANNEL_FMT_ONLY_RIGHT,
//       .communication_format = I2S_COMM_FORMAT_I2S,
//       .intr_alloc_flags = 0,
//       .dma_buf_count = 8,
//       .dma_buf_len = 512,
//       .use_apll = false,
//       .tx_desc_auto_clear = false,
//       .fixed_mclk = -1,
//   };
//   i2s_pin_config_t pin_config = {
//       .bck_io_num = 26,    // IIS_SCLK
//       .ws_io_num = 32,     // IIS_LCLK
//       .data_out_num = -1,  // IIS_DSIN
//       .data_in_num = 33,   // IIS_DOUT
//   };
//   esp_err_t ret = 0;

//   ret = i2s_driver_install((i2s_port_t)1, &i2s_config, 0, NULL);
//   if (ret != ESP_OK) {
//     ei_printf("Error in i2s_driver_install");
//   }

//   ret = i2s_set_pin((i2s_port_t)1, &pin_config);
//   if (ret != ESP_OK) {
//     ei_printf("Error in i2s_set_pin");
//   }

//   ret = i2s_zero_dma_buffer((i2s_port_t)1);
//   if (ret != ESP_OK) {
//     ei_printf("Error in initializing dma buffer with 0");
//   }

//   return int(ret);
// }

// static int i2s_deinit(void) {
//     i2s_driver_uninstall((i2s_port_t)1); //stop & destroy i2s driver
//     return 0;
// }

#if !defined(EI_CLASSIFIER_SENSOR) || EI_CLASSIFIER_SENSOR != EI_CLASSIFIER_SENSOR_MICROPHONE
#error "Invalid model for current sensor."
#endif
</font>
复制代码

输出该信息对应的代码部分为

对该段代码的变量说明：
result.classification[ix].label是标签名（noise,off,on）；
result.classification[ix].value是当前声音为三个标签的可能性；
pred_index是最大可能标签的序号，noise--0,off--1,on--2.

                    
上两个截图是表示当时麦克风听到“关灯”“开灯”的方言词。

上面的代码是根据听到的命令（标签值）控制板载LED的亮与灭。

全文完