feat(system): Support ESP32

CMakeLists.txt

@@ -15,16 +15,31 @@ set(COMPONENT_REQUIRES
 register_component()
 if(IDF_TARGET STREQUAL "esp32")
     target_link_libraries(${COMPONENT_TARGET} "-L ${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32")
+    target_link_libraries(${COMPONENT_TARGET} "-L ${CMAKE_CURRENT_SOURCE_DIR}/esp-tts/esp_tts_chinese")
+
     add_prebuilt_library(esp_audio_processor "${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32/libesp_audio_processor.a" PRIV_REQUIRES esp-sr)
     add_prebuilt_library(wakenet "${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32/libwakenet.a" PRIV_REQUIRES esp-sr)
     add_prebuilt_library(multinet "${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32/libmultinet.a" PRIV_REQUIRES esp-sr)
     add_prebuilt_library(esp_audio_front_end "${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32/libesp_audio_front_end.a" PRIV_REQUIRES esp-sr)
-    add_prebuilt_library(hufzip "${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32/libhufzip.a" PRIV_REQUIRES esp-sr)
-    # TTS
     target_link_libraries(${COMPONENT_TARGET} "-Wl,--start-group"
-        esp_tts_chinese_esp32s2 
-        voice_set_xiaole_esp32s2
-        voice_set_template_esp32s2
+        wakenet
+        dl_lib
+        c_speech_features
+        hilexin_wn5
+        hilexin_wn5X2
+        hilexin_wn5X3
+        nihaoxiaozhi_wn5
+        nihaoxiaozhi_wn5X2
+        nihaoxiaozhi_wn5X3
+        nihaoxiaoxin_wn5X3
+        customized_word_wn5
+        multinet
+        multinet2_ch
+        esp_audio_processor 
+        esp_audio_front_end 
+        esp_tts_chinese 
+        voice_set_xiaole 
+        voice_set_template
         "-Wl,--end-group")
 elseif(${IDF_TARGET} STREQUAL "esp32s2")
     target_link_libraries(${COMPONENT_TARGET} "-Wl,--start-group"
@@ -34,6 +49,7 @@ elseif(${IDF_TARGET} STREQUAL "esp32s2")
          "-Wl,--end-group")
 elseif(${IDF_TARGET} STREQUAL "esp32s3")
     target_link_libraries(${COMPONENT_TARGET} "-L ${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32s3")
+
     add_prebuilt_library(esp_audio_processor "${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32s3/libesp_audio_processor.a" PRIV_REQUIRES esp-sr)
     add_prebuilt_library(wakenet "${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32s3/libwakenet.a" PRIV_REQUIRES esp-sr)
     add_prebuilt_library(multinet "${CMAKE_CURRENT_SOURCE_DIR}/lib/esp32s3/libmultinet.a" PRIV_REQUIRES esp-sr)

Kconfig.projbuild

@@ -27,6 +27,7 @@ endchoice
 choice MODEL_DATA_PATH
     prompt "model data path"
     default MODEL_IN_SPIFFS
+    depends on IDF_TARGET_ESP32S3
 
     config MODEL_IN_SPIFFS
         bool "spiffs partition"
@@ -103,10 +104,6 @@ choice SR_WN_WAKE_WORD_SEL
         bool "nihaoxiaozhi (WakeNet5X3)"
         depends on SR_WN_MODEL_WN5_QUANT || SR_WN_MODEL_WN5_FLOAT
 
-    config SR_WN_WN5X3_HIJESON
-        bool "hi jeson (WakeNet5X3)"
-        depends on SR_WN_MODEL_WN5_QUANT || SR_WN_MODEL_WN5_FLOAT 
-
     config SR_WN_WN5X3_NIHAOXIAOXIN
         bool "nihaoxiaoxin (WakeNet5X3)"
         depends on SR_WN_MODEL_WN5_QUANT || SR_WN_MODEL_WN5_FLOAT 
@@ -157,6 +154,7 @@ choice SR_MN_LANGUAGE_SEL
 
     config SR_MN_ENGLISH
         bool "english"
+        depends on IDF_TARGET_ESP32S3
 
 endchoice
 
@@ -187,10 +185,6 @@ choice SR_MN_MODE_SEL
         bool "chinese single recognition (MultiNet2)"
         depends on SR_MN_CHINESE && IDF_TARGET_ESP32
 
-    config SR_MN_CN_MULTINET2_CONTINUOUS_RECOGNITION
-        bool "chinese continuous recognition (MultiNet2)"
-        depends on SR_MN_CHINESE && IDF_TARGET_ESP32
-
     config SR_MN_CN_MULTINET3_SINGLE_RECOGNITION
         bool "chinese single recognition (MultiNet3)"
         depends on SR_MN_CHINESE && IDF_TARGET_ESP32S3

include/customized_word_wn5.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_customized_word_wn5;

include/dl_lib.h

@@ -20,12 +20,11 @@
 
 typedef int padding_state;
 
-#ifdef CONFIG_IDF_TARGET_ESP32S3
 void *dl_lib_calloc(int cnt, int size, int align);
+
 void *dl_lib_calloc_psram(int cnt, int size, int align);
 
 void dl_lib_free(void *d);
-#endif
 /**
  * @brief Does a fast version of the exp() operation on a floating point number.
  *

include/esp_afe_sr_iface.h

@@ -41,6 +41,25 @@ typedef struct {
     int agc_mode;
 } afe_config_t;
 
+
+#if CONFIG_IDF_TARGET_ESP32
+#define AFE_CONFIG_DEFAULT() { \
+    .aec_init = true, \
+    .se_init = true, \
+    .vad_init = true, \
+    .wakenet_init = true, \
+    .vad_mode = 3, \
+    .wakenet_model = &WAKENET_MODEL, \
+    .wakenet_coeff = &WAKENET_COEFF, \
+    .wakenet_mode = DET_MODE_90, \
+    .afe_mode = SR_MODE_HIGH_PERF, \
+    .afe_perferred_core = 0, \
+    .afe_perferred_priority = 5, \
+    .afe_ringbuf_size = 50, \
+    .alloc_from_psram = 1, \
+    .agc_mode = 2, \
+}
+#elif CONFIG_IDF_TARGET_ESP32S3
 #define AFE_CONFIG_DEFAULT() { \
     .aec_init = true, \
     .se_init = true, \
@@ -57,7 +76,7 @@ typedef struct {
     .alloc_from_psram = 1, \
     .agc_mode = 2, \
 }
-
+#endif
 /**
  * @brief Function to initialze a AFE_SR instance with a specified mode
  * 

include/hijeson_wn5X3.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_hijeson_wn5X3;

include/hilexin_wn5.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_hilexin_wn5;

include/hilexin_wn5X2.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_hilexin_wn5X2;

include/hilexin_wn5X3.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_hilexin_wn5X3;

include/multinet2_ch.h

@@ -0,0 +1,9 @@
+//Generated by mkmodel_py
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+#include "dl_lib_matrixq8.h"
+
+extern const model_coeff_getter_t get_coeff_multinet2_ch;

include/nihaoxiaoxin_wn5X3.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_nihaoxiaoxin_wn5X3;

include/nihaoxiaoxin_wn6.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_nihaoxiaoxin_wn6;

include/nihaoxiaozhi_wn5.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_nihaoxiaozhi_wn5;

include/nihaoxiaozhi_wn5X2.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_nihaoxiaozhi_wn5X2;

include/nihaoxiaozhi_wn5X3.h

@@ -0,0 +1,8 @@
+//Generated by mkmodel
+#pragma once
+#include <string.h>
+#include "dl_lib_coefgetter_if.h"
+#include "dl_lib_matrix.h"
+#include "dl_lib_matrixq.h"
+
+extern const model_coeff_getter_t get_coeff_nihaoxiaozhi_wn5X3;

libversion

@@ -1 +1 @@
-d187cb93b72cefcef1904550d50c5f1868bb0099
+695cd3da8e08da040b6f69bce2e55c3f3c77a902

src/esp_afe_sr_1mic.c

@@ -95,14 +95,20 @@ static esp_afe_sr_data_t *afe_create_from_config(afe_config_t *afe_config)
         }
 
         if (afe->mode == SR_MODE_LOW_COST) {
+#ifdef CONFIG_IDF_TARGET_ESP32S3
             afe->aec_handle = esp_aec3_728_init(afe->nch, afe->aec_frame_size,
                                                 afe->aec_filter_length, afe_config->alloc_from_psram);
+#else
+            printf("Not support SR_MODE_LOW_COST on ESP32\n");
+            afe->aec_handle = NULL;
+#endif
         } else {
             afe->aec_handle = esp_aec3_init(afe->nch, afe->aec_frame_size,
-                                            afe->aec_filter_length, 1);
+                                            afe->aec_filter_length, 0);
         }
     } else {
         afe->aec_frame_size = 512;
+        afe->aec_nch = afe->nch + 1;
         afe->aec_handle = NULL;
     }
 
@@ -159,7 +165,11 @@ static esp_afe_sr_data_t *afe_create_from_config(afe_config_t *afe_config)
         afe->wn_gain = 2;
         afe->agc_mode = 0;
     }
+    if (afe_config->alloc_from_psram) {
+        afe->buff_wn = heap_caps_malloc(afe->audio_chunksize * (afe->wn_nch + afe->wn_nch - afe->nch) * sizeof(int16_t), MALLOC_CAP_SPIRAM);
+    } else {
         afe->buff_wn = malloc(afe->audio_chunksize * (afe->wn_nch + afe->wn_nch - afe->nch) * sizeof(int16_t));
+    }
 
     /********************************************************/
     /**********************INIT OTRHERS**********************/
@@ -209,8 +219,10 @@ static int afe_feed(esp_afe_sr_data_t *afe, int16_t *in)
         if (frame_cnt++ < 4) ref_vad_res = 0;
 
         if (afe->mode == SR_MODE_LOW_COST) {
+#ifdef CONFIG_IDF_TARGET_ESP32S3
             esp_aec3_728_write_ref_vad(afe->aec_handle, ref_vad_res);
             esp_aec3_728_process(afe->aec_handle, afe->aec_in, afe->aec_in + afe->aec_frame_size, afe->aec_tmp);
+#endif
         } else if (afe->mode == SR_MODE_HIGH_PERF) {
             esp_aec3_write_ref_vad(afe->aec_handle, ref_vad_res);
             esp_aec3_process(afe->aec_handle, afe->aec_in, afe->aec_in + afe->aec_frame_size, afe->aec_tmp);
@@ -266,7 +278,9 @@ static void afe_se_task(void *arg)
                 }
             }
         }
-
+        // if (sr_rb_available(afe->rb_out) < afe->wn_nch * afe->ns_frame_size * sizeof(int16_t)) {
+        //     printf("ERROR! rb_out slow!!!\n");
+        // }
         sr_rb_write(afe->rb_out, afe->ns_in, afe->wn_nch * afe->ns_frame_size * sizeof(int16_t), 0);
     }
     printf("afe_se_task quit\n");
@@ -302,9 +316,9 @@ static int afe_fetch(esp_afe_sr_data_t *afe, int16_t *out)
         res = afe->wakenet->detect(afe->model_data, afe->buff_wn);
         //selector & gainer
         afe->channel_id = afe->wakenet->get_triggered_channel(afe->model_data);
-        if (res > 0 && afe->agc_mode>=0) {
+        if (res > 0 && afe->agc_mode >= 0) {
             if (afe->agc_mode == 0) out_gain = 1;
-            else out_gain = afe->wakenet->get_vol_gain(afe->model_data, (afe->agc_mode-6));
+            else out_gain = afe->wakenet->get_vol_gain(afe->model_data, (afe->agc_mode - 6));
         }
 
         int shift = afe->audio_chunksize * afe->channel_id;
@@ -435,8 +449,10 @@ static void afe_destory(esp_afe_sr_data_t *afe)
 
         if (afe->aec_handle != NULL) {
             if (afe->mode == SR_MODE_LOW_COST) {
+#ifdef CONFIG_IDF_TARGET_ESP32S3
                 vad_destroy(afe->ref_vad_handle);
                 esp_aec3_728_destroy(afe->aec_handle);
+#endif
             }
             else if (afe->mode == SR_MODE_HIGH_PERF) {
                 vad_destroy(afe->ref_vad_handle);
@@ -454,11 +470,11 @@ static void afe_destory(esp_afe_sr_data_t *afe)
 
         if (afe->wakenet != NULL) {
             afe->wakenet->destroy(afe->model_data);
+            afe->wakenet = NULL;
+        }
         if (afe->buff_wn != NULL) {
             free(afe->buff_wn);
-            }
-
-            afe->wakenet = NULL;
+            afe->buff_wn = NULL;
         }
 
         if (afe->vad_handle != NULL) {