Update sr.

.gitignore

@@ -1,4 +1,3 @@
 build
 managed_components
-*.lock
-sdkconfig
+*.lock

.vscode/settings.json

@@ -46,6 +46,7 @@
     "stdbool.h": "c",
     "esp_afe_sr_models.h": "c",
     "esp_afe_config.h": "c",
-    "esp_afe_sr_iface.h": "c"
+    "esp_afe_sr_iface.h": "c",
+    "esp_err.h": "c"
   }
 }

main/app_sr.c

@@ -7,6 +7,7 @@
 // #include "audio_player.h"
 // #include "esp32_s3_szp.h"
 
+#include "dl_lib_coefgetter_if.h"
 #include "esp_afe_sr_iface.h"
 #include "esp_afe_sr_models.h"
 #include "esp_err.h"
@@ -14,6 +15,7 @@
 #include "esp_mn_iface.h"
 #include "esp_mn_models.h"
 #include "esp_mn_speech_commands.h"
+#include "esp_nsn_models.h"
 #include "esp_process_sdkconfig.h"
 #include "esp_wn_iface.h"
 #include "esp_wn_models.h"
@@ -57,7 +59,7 @@ esp_err_t record_start2() {
   esp_err_t ret = ESP_OK;
 
   const wav_header_t wav_header2 =
-      WAV_HEADER_PCM_DEFAULT(16000 * 1 * 16 / 8 * 20, 16, 16000, 1);
+      WAV_HEADER_PCM_DEFAULT(16000 * 2 * 16 / 8 * 20, 16, 16000, 2);
 
   f2 = fopen("/sdcard/bbbbb.wav", "w");
   ESP_LOGI(TAG, "fopen error: %s", strerror(errno));
@@ -78,6 +80,7 @@ esp_err_t record_add1(int16_t* data1, uint32_t size) {
     // ESP_RETURN_ON_FALSE(fwrite(data1, size, 1, f1), ESP_FAIL, TAG,
     //                     "error while writing samples to wav file");
     wav_written1 += size;
+    // ESP_LOGI(TAG, "Recording %d !", wav_written1);
     return ESP_OK;
   } else {
     if (wav_written1 < 0xFFFFFFFF) {
@@ -93,7 +96,7 @@ bool record_end1() { return wav_written1 == 0xFFFFFFFF; }
 bool record_end2() { return wav_written2 == 0xFFFFFFFF; }
 
 esp_err_t record_add2(int16_t* data2, uint32_t size) {
-  uint32_t byte_rate = 16000 * 1 * 16 / 8;
+  uint32_t byte_rate = 16000 * 2 * 16 / 8;
   uint32_t wav_size = byte_rate * 20;
 
   if (wav_written2 < wav_size) {
@@ -112,9 +115,30 @@ esp_err_t record_add2(int16_t* data2, uint32_t size) {
   return ESP_OK;
 }
 
-void afe(esp_afe_sr_data_t* afe_data, int16_t* i2s_buff) {
-  afe_handle->feed(afe_data, i2s_buff);  // 把获取到的I2S数据输入给afe_data
-  afe_fetch_result_t* res = afe_handle->fetch(afe_data);
+uint32_t afe_pipe = 0;
+
+afe_fetch_result_t* afe(esp_afe_sr_data_t* afe_data, int16_t* i2s_buff) {
+  afe_handle->feed(afe_data, i2s_buff);
+
+  // printf("AFE get_feed_chunksize: %d\n",
+  //        afe_handle->get_feed_chunksize(afe_data));
+  // printf("AFE get_fetch_chunksize: %d\n",
+  //        afe_handle->get_fetch_chunksize(afe_data));
+  // afe_handle->print_pipeline(afe_data);
+  // printf("AFE get_channel_num: %d\n", afe_handle->get_channel_num(afe_data));
+  // printf("AFE get_samp_rate: %d\n", afe_handle->get_samp_rate(afe_data));
+
+  afe_fetch_result_t* res = NULL;
+  if (afe_pipe >= 5) {
+    res = afe_handle->fetch(afe_data);
+    // printf("AFE data_size: %d\n", res->data_size);
+    // printf("AFE data_volume: %f\n", res->data_volume);
+    // printf("AFE vad_state: %d\n", res->vad_state);
+    // printf("AFE ret_value: %d\n", res->ret_value);
+  } else {
+    afe_pipe++;
+  }
+  return res;
 }
 
 void feed_Task(void* arg) {
@@ -288,82 +312,13 @@ void detect_Task(void* arg) {
 }
 
 esp_afe_sr_data_t* sr_init() {
-  // 获取模型 名称“model”和分区表中装载模型的名称一致
-  models = esp_srmodel_init("model");
-
-  afe_handle = (esp_afe_sr_iface_t*)&ESP_AFE_SR_HANDLE;
-  // 先配置afe句柄
-  // 随后才可以调用afe接口
-  // afe_config_t afe_config = AFE_CONFIG_DEFAULT();  // 配置afe
-  afe_config_t afe_config = {
-      .aec_init = true,
-      .se_init = true,
-      .vad_init = true,
-      .wakenet_init = true,
-      .voice_communication_init = false,
-      .voice_communication_agc_init = false,
-      .voice_communication_agc_gain = 15,
-      .vad_mode = VAD_MODE_3,
-      .wakenet_model_name = NULL,
-      .wakenet_model_name_2 = NULL,
-      .wakenet_mode = DET_MODE_2CH_90,
-      .afe_mode = SR_MODE_LOW_COST,
-      .afe_perferred_core = 0,
-      .afe_perferred_priority = 5,
-      .afe_ringbuf_size = 50,
-      .memory_alloc_mode = AFE_MEMORY_ALLOC_MORE_PSRAM,
-      .afe_linear_gain = 1.0,
-      .agc_mode = AFE_MN_PEAK_AGC_MODE_2,
-      .pcm_config =
-          {
-              .total_ch_num = 3,
-              .mic_num = 2,
-              .ref_num = 1,
-              .sample_rate = 16000,
-          },
-      .debug_init = false,
-      .debug_hook = {{AFE_DEBUG_HOOK_MASE_TASK_IN, NULL},
-                     {AFE_DEBUG_HOOK_FETCH_TASK_IN, NULL}},
-      .afe_ns_mode = NS_MODE_SSP,
-      .afe_ns_model_name = NULL,
-      .fixed_first_channel = true,
-  };
-
-  // 配置唤醒模型 必须在create_from_config之前配置
-
-  afe_config.aec_init = true;
-  afe_config.se_init = true;
-  afe_config.vad_init = true;
-  afe_config.wakenet_init = false;
-  afe_config.voice_communication_init = false;
-  afe_config.voice_communication_agc_init = true;
-  afe_config.voice_communication_agc_gain = 15;
-  afe_config.vad_mode = VAD_MODE_3;
-  afe_config.wakenet_model_name = NULL;
-  afe_config.wakenet_model_name_2 = NULL;
-  afe_config.wakenet_mode = DET_MODE_2CH_90;
-  afe_config.afe_mode = SR_MODE_LOW_COST;
-  afe_config.afe_perferred_core = 0;
-  afe_config.afe_perferred_priority = 5;
-  afe_config.afe_ringbuf_size = 50;
-  afe_config.memory_alloc_mode = AFE_MEMORY_ALLOC_MORE_PSRAM;
-  afe_config.afe_linear_gain = 4.0;
-  afe_config.agc_mode = AFE_MN_PEAK_AGC_MODE_2;
-  afe_config.pcm_config.total_ch_num = 3;
-  afe_config.pcm_config.mic_num = 2;
-  afe_config.pcm_config.ref_num = 1;
-  afe_config.pcm_config.sample_rate = 16000;
-  afe_config.debug_init = false;
-  // afe_config.debug_hook = {{AFE_DEBUG_HOOK_MASE_TASK_IN, NULL},
-  //                          {AFE_DEBUG_HOOK_FETCH_TASK_IN, NULL}};
-  afe_config.afe_ns_mode = NS_MODE_SSP;
-  afe_config.afe_ns_model_name = NULL;
-
-  // afe_config.wakenet_model_name =
-  //     esp_srmodel_filter(models, ESP_WN_PREFIX, NULL);
-  return afe_handle->create_from_config(&afe_config);  // 创建afe_data
-  // ESP_LOGI(TAG, "wakenet:%s", afe_config.wakenet_model_name);  //
-  // 打印唤醒名称
+  srmodel_list_t* models = esp_srmodel_init("model");
+  afe_config_t* afe_config =
+      afe_config_init("MM", models, AFE_TYPE_VC, AFE_MODE_LOW_COST);
+  afe_handle = esp_afe_handle_from_config(afe_config);
+  esp_afe_sr_data_t* afe_data = afe_handle->create_from_config(afe_config);
+  afe_config_free(afe_config);
+  return afe_data;
 }
 
 void app_sr_init(void) {

main/app_sr.h

@@ -6,11 +6,9 @@
 #include "esp_afe_sr_models.h"
 #include "esp_err.h"
 
-void app_sr_init(void);
+esp_afe_sr_data_t* sr_init();
 
-void app_sr_init(void);
-
-void afe(esp_afe_sr_data_t* afe_data, int16_t* i2s_buff);
+afe_fetch_result_t* afe(esp_afe_sr_data_t* afe_data, int16_t* i2s_buff);
 
 bool record_end1();
 bool record_end2();

main/idf_component.yml

@@ -4,7 +4,7 @@ dependencies:
     version: "^2.0.0"
   espressif/es7210: "^1.0.0"
   chmorgan/esp-file-iterator: "1.0.0"       # 获取文件
-  espressif/esp-sr: "^1.6.0"                # 语音识别
+  espressif/esp-sr: "^2.0.0"                # 语音识别
   idf:
     version: ">=5.4.0"
 

main/main.c

@@ -69,7 +69,7 @@
 #define SAMPLE_RATE 16000
 #define I2S_PORT I2S_NUM_0
 #define DMA_BUF_COUNT 8
-#define DMA_BUF_LEN 1024
+#define DMA_BUF_LEN 512
 #define SAMPLE_BITS I2S_DATA_BIT_WIDTH_16BIT
 
 // 引脚定义
@@ -82,6 +82,8 @@
 static const char* TAG = "I2S_DMA_ISR";
 static i2s_chan_handle_t rx_chan = NULL;
 
+esp_afe_sr_data_t* afe_sr_data = NULL;
+
 // 中断服务程序的数据处理任务句柄
 static TaskHandle_t rx_task_handle = NULL;
 
@@ -247,9 +249,22 @@ static void i2s_rx_task(void* arg) {
       // }
 
       record_add1(data_buffer, bytes_read);
-      if (!record_end1()) ESP_LOGI(TAG, "处理 %d 字节的I2S数据", bytes_read);
+      // if (!record_end1()) ESP_LOGI(TAG, "处理 %d 字节的I2S数据", bytes_read);
 
-      vTaskDelay(10 / portTICK_PERIOD_MS);
+      afe_fetch_result_t* res = afe(afe_sr_data, data_buffer);
+      if (res != NULL) {
+        record_add2(res->data, res->data_size);
+        if (res->vad_state > 0) {
+          printf("AFE data_size: %d\n", res->data_size);
+          // printf("AFE data_volume: %f\n", res->data_volume);
+          // printf("AFE vad_state: %d\n", res->vad_state);
+          // printf("AFE ret_value: %d\n", res->ret_value);
+        }
+      }
+
+      if (record_end1() && record_end2()) break;
+
+      // vTaskDelay(10 / portTICK_PERIOD_MS);
 
       // esp_sleep_enable_timer_wakeup(50 * 1000);
       // esp_light_sleep_start();
@@ -407,11 +422,12 @@ void app_main(void) {
 
   ESP_LOGI(TAG, "启动I2S DMA接收示例");
 
-  // app_sr_init();  // 语音识别初始化
+  afe_sr_data = sr_init();
 
   config_power_manager();
 
   record_start1();
+  record_start2();
 
   BaseType_t task_ret = xTaskCreate(i2s_rx_task, "i2s_rx_task", 4096, NULL,
                                     tskIDLE_PRIORITY + 2, &rx_task_handle);
@@ -429,7 +445,7 @@ void app_main(void) {
   //   vTaskDelay(pdMS_TO_TICKS(1000));
   // }
 
-  while (!record_end1()) {
+  while (!record_end1() || !record_end2()) {
     vTaskDelay(pdMS_TO_TICKS(1000));
   }
 

sdkconfig.defaults

@@ -67,10 +67,11 @@ CONFIG_MODEL_IN_FLASH=y
 # CONFIG_MODEL_IN_SDCARD is not set
 CONFIG_USE_AFE=y
 CONFIG_AFE_INTERFACE_V1=y
-# CONFIG_USE_NSNET is not set
-CONFIG_USE_WAKENET=y
+CONFIG_USE_NSNET=y
+CONFIG_SR_NSN_NSNET2=y
+# CONFIG_USE_WAKENET=y
 # CONFIG_SR_WN_WN8_ALEXA is not set
-CONFIG_SR_WN_WN9_HILEXIN=y
+# CONFIG_SR_WN_WN9_HILEXIN=y
 # CONFIG_SR_WN_WN9_XIAOAITONGXUE is not set
 # CONFIG_SR_WN_WN9_ALEXA is not set
 # CONFIG_SR_WN_WN9_HIESP is not set
@@ -81,14 +82,14 @@ CONFIG_SR_WN_WN9_HILEXIN=y
 # CONFIG_SR_WN_WN9_HEYWILLOW_TTS is not set
 # CONFIG_SR_WN_WN9_CUSTOMWORD is not set
 # CONFIG_SR_WN_LOAD_MULIT_WORD is not set
-CONFIG_USE_MULTINET=y
+# CONFIG_USE_MULTINET=y
 # CONFIG_SR_MN_CN_NONE is not set
 # CONFIG_SR_MN_CN_MULTINET5_RECOGNITION_QUANT8 is not set
-CONFIG_SR_MN_CN_MULTINET6_QUANT=y
+# CONFIG_SR_MN_CN_MULTINET6_QUANT=y
 # CONFIG_SR_MN_CN_MULTINET6_AC_QUANT is not set
 # CONFIG_SR_MN_CN_MULTINET7_QUANT is not set
 # CONFIG_SR_MN_CN_MULTINET7_AC_QUANT is not set
-CONFIG_SR_MN_EN_NONE=y
+# CONFIG_SR_MN_EN_NONE=y
 # CONFIG_SR_MN_EN_MULTINET5_SINGLE_RECOGNITION_QUANT8 is not set
 # CONFIG_SR_MN_EN_MULTINET6_QUANT is not set
 # CONFIG_SR_MN_EN_MULTINET7_QUANT is not set