feat(AFE): modify the API of AFE to support multiple version

Kconfig.projbuild

@@ -12,6 +12,24 @@ choice MODEL_DATA_PATH
         bool "SD Card"
 endchoice
 
+
+config USE_AFE
+	bool "use afe"
+	default "y"
+
+choice AFE_INTERFACE_SEL
+	prompt "Afe interface selection"
+	default AFE_INTERFACE_V1
+	depends on USE_AFE
+	help
+		Select the afe interface to be used.
+
+	config AFE_INTERFACE_V1
+		bool "afe interface v1"
+
+endchoice
+
+
 config USE_WAKENET
     bool "use wakenet"
     default "y"

include/esp32s3/esp_afe_config.h

@@ -0,0 +1,98 @@
+#pragma once
+#include "stdint.h"
+#include "esp_wn_iface.h"
+#include "esp_wn_models.h"
+#include "esp_vad.h"
+
+//AFE: Audio Front-End 
+//SR:  Speech Recognition
+//afe_sr/AFE_SR: the audio front-end for speech recognition
+
+//Set AFE_SR mode
+typedef enum {
+    SR_MODE_LOW_COST = 0,
+    SR_MODE_HIGH_PERF = 1
+} afe_sr_mode_t;
+
+typedef enum {
+    AFE_MEMORY_ALLOC_MORE_INTERNAL = 0,             // malloc with more internal ram
+    AFE_MEMORY_ALLOC_INTERNAL_PSRAM_BALANCE = 1,    // malloc with internal ram and psram in balance
+    AFE_MEMORY_ALLOC_MORE_PSRAM = 2                 // malloc with more psram
+} afe_memory_alloc_mode_t;
+
+typedef enum {
+    AFE_MN_PEAK_AGC_MODE_1 = -5,            // The peak amplitude of audio fed to multinet is -5dB
+    AFE_MN_PEAK_AGC_MODE_2 = -4,            // The peak amplitude of audio fed to multinet is -4dB
+    AFE_MN_PEAK_AGC_MODE_3 = -3,            // The peak amplitude of audio fed to multinet is -3dB
+    AFE_MN_PEAK_NO_AGC = 0,                 // There is no agc gain
+} afe_mn_peak_agc_mode_t;
+
+typedef struct {
+    int total_ch_num;                       // total channel num. It must be: total_ch_num = mic_num + ref_num
+    int mic_num;                            // mic channel num
+    int ref_num;                            // reference channel num
+} afe_pcm_config_t;
+
+typedef struct {
+    bool aec_init;
+    bool se_init;
+    bool vad_init;
+    bool wakenet_init;
+    bool voice_communication_init;
+    vad_mode_t vad_mode;                    // The value can be: VAD_MODE_0, VAD_MODE_1, VAD_MODE_2, VAD_MODE_3, VAD_MODE_4
+    esp_wn_iface_t *wakenet_model;
+    model_coeff_getter_t *wakenet_coeff;
+    det_mode_t wakenet_mode;
+    afe_sr_mode_t afe_mode;
+    int afe_perferred_core;
+    int afe_perferred_priority;
+    int afe_ringbuf_size;
+    afe_memory_alloc_mode_t memory_alloc_mode;
+    afe_mn_peak_agc_mode_t agc_mode;
+    afe_pcm_config_t pcm_config;            // Config the channel num of original data which is fed to the afe feed function.
+} afe_config_t;
+
+
+#if CONFIG_IDF_TARGET_ESP32
+#define AFE_CONFIG_DEFAULT() { \
+    .aec_init = true, \
+    .se_init = true, \
+    .vad_init = true, \
+    .wakenet_init = true, \
+    .voice_communication_init = false, \
+    .vad_mode = 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, \
+    .memory_alloc_mode = AFE_MEMORY_ALLOC_INTERNAL_PSRAM_BALANCE, \
+    .agc_mode = AFE_MN_PEAK_AGC_MODE_2, \
+    .pcm_config.total_ch_num = 2, \
+    .pcm_config.mic_num = 1, \
+    .pcm_config.ref_num = 1, \
+}
+#elif CONFIG_IDF_TARGET_ESP32S3
+#define AFE_CONFIG_DEFAULT() { \
+    .aec_init = true, \
+    .se_init = true, \
+    .vad_init = true, \
+    .wakenet_init = true, \
+    .voice_communication_init = false, \
+    .vad_mode = VAD_MODE_3, \
+    .wakenet_model = &WAKENET_MODEL, \
+    .wakenet_coeff = &WAKENET_COEFF, \
+    .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, \
+    .agc_mode = AFE_MN_PEAK_AGC_MODE_2, \
+    .pcm_config.total_ch_num = 3, \
+    .pcm_config.mic_num = 2, \
+    .pcm_config.ref_num = 1, \
+}
+#endif

include/esp32s3/esp_afe_sr_iface.h

@@ -1,8 +1,10 @@
 #pragma once
 #include "stdint.h"
-#include "esp_wn_iface.h"
-#include "esp_wn_models.h"
-#include "esp_vad.h"
+#if CONFIG_AFE_INTERFACE_V1
+#include "esp_afe_config.h"
+#else
+#include "esp_afe_config.h"
+#endif
 
 //AFE: Audio Front-End 
 //SR:  Speech Recognition
@@ -11,12 +13,6 @@
 //Opaque AFE_SR data container
 typedef struct esp_afe_sr_data_t esp_afe_sr_data_t;
 
-//Set AFE_SR mode
-typedef enum {
-    SR_MODE_LOW_COST = 0,
-    SR_MODE_HIGH_PERF = 1
-} afe_sr_mode_t;
-
 // the output state of fetch function
 typedef enum {
     AFE_FETCH_ERROR = -3,                   // fetch empty data, retry it
@@ -26,85 +22,6 @@ typedef enum {
     AFE_FETCH_WWE_DETECTED = 1              // wwe state: wake word is detected
 } afe_fetch_mode_t;
 
-typedef enum {
-    AFE_PSRAM_LOW_COST = 0,
-    AFE_PSRAM_MIDDLE_COST = 1,
-    AFE_PSRAM_HIGH_COST = 2
-} afe_use_psram_mode_t;
-
-typedef enum {
-    AFE_MN_PEAK_AGC_MODE_1 = -5,            // The peak amplitude of audio fed to multinet is -5dB
-    AFE_MN_PEAK_AGC_MODE_2 = -4,            // The peak amplitude of audio fed to multinet is -4dB
-    AFE_MN_PEAK_AGC_MODE_3 = -3,            // The peak amplitude of audio fed to multinet is -3dB
-    AFE_MN_PEAK_NO_AGC = 0,                 // There is no agc gain
-} afe_mn_peak_agc_mode_t;
-
-typedef struct {
-    int total_ch_num;                       // total channel num. It must be: total_ch_num = mic_num + ref_num
-    int mic_num;                            // mic channel num
-    int ref_num;                            // reference channel num
-} afe_pcm_config_t;
-
-typedef struct {
-    bool aec_init;
-    bool se_init;
-    bool vad_init;
-    bool wakenet_init;
-    vad_mode_t vad_mode;                    // The value can be: VAD_MODE_0, VAD_MODE_1, VAD_MODE_2, VAD_MODE_3, VAD_MODE_4
-    esp_wn_iface_t *wakenet_model;
-    model_coeff_getter_t *wakenet_coeff;
-    det_mode_t wakenet_mode;
-    afe_sr_mode_t afe_mode;
-    int afe_perferred_core;
-    int afe_perferred_priority;
-    int afe_ringbuf_size;
-    afe_use_psram_mode_t alloc_from_psram;
-    afe_mn_peak_agc_mode_t agc_mode;
-    afe_pcm_config_t pcm_config;            // Config the channel num of original data which is fed to the afe feed function.
-} 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 = 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 = AFE_PSRAM_MIDDLE_COST, \
-    .agc_mode = AFE_MN_PEAK_AGC_MODE_2, \
-    .pcm_config.total_ch_num = 2, \
-    .pcm_config.mic_num = 1, \
-    .pcm_config.ref_num = 1, \
-}
-#elif CONFIG_IDF_TARGET_ESP32S3
-#define AFE_CONFIG_DEFAULT() { \
-    .aec_init = true, \
-    .se_init = true, \
-    .vad_init = true, \
-    .wakenet_init = true, \
-    .vad_mode = VAD_MODE_3, \
-    .wakenet_model = &WAKENET_MODEL, \
-    .wakenet_coeff = &WAKENET_COEFF, \
-    .wakenet_mode = DET_MODE_2CH_90, \
-    .afe_mode = SR_MODE_LOW_COST, \
-    .afe_perferred_core = 0, \
-    .afe_perferred_priority = 5, \
-    .afe_ringbuf_size = 50, \
-    .alloc_from_psram = AFE_PSRAM_HIGH_COST, \
-    .agc_mode = AFE_MN_PEAK_AGC_MODE_2, \
-    .pcm_config.total_ch_num = 3, \
-    .pcm_config.mic_num = 2, \
-    .pcm_config.ref_num = 1, \
-}
-#endif
 /**
  * @brief Function to initialze a AFE_SR instance with a specified mode
  * 
@@ -274,5 +191,3 @@ typedef struct {
     esp_afe_sr_iface_op_enable_se_t enable_se;
     esp_afe_sr_iface_op_destroy_t destroy;
 } esp_afe_sr_iface_t;
-
-extern esp_afe_sr_iface_t esp_afe_sr;

include/esp32s3/esp_afe_sr_models.h

@@ -1,6 +1,18 @@
 #pragma once
+
+#if defined CONFIG_USE_AFE
+
+#if CONFIG_AFE_INTERFACE_V1
 #include "esp_afe_sr_iface.h"
+extern const esp_afe_sr_iface_t esp_afe_v1;
+#define ESP_AFE_HANDLE esp_afe_v1
+#else
+#error No valid afe selected.
+#endif
 
-extern const esp_afe_sr_iface_t esp_afe_sr_2mic;
-extern const esp_afe_sr_iface_t esp_afe_sr_1mic;
+#else
+#include "esp_afe_sr_iface.h"
+extern const esp_afe_sr_iface_t esp_afe_v1;
+#define ESP_AFE_HANDLE esp_afe_v1
 
+#endif

include/esp32s3/esp_agc.h

@@ -14,10 +14,6 @@
 #ifndef _ESP_AGC_H_
 #define _ESP_AGC_H_
 
-#ifdef __cplusplus
-extern "C" {
-#endif
-
 ////all positive value is valid, negective is error
 typedef enum {
     ESP_AGC_SUCCESS = 0,   ////success
@@ -32,8 +28,4 @@ void set_agc_config(void *agc_handle, int gain_dB, int limiter_enable, int targe
 int esp_agc_process(void *agc_handle, short *in_pcm, short *out_pcm, int frame_size, int sample_rate);
 void esp_agc_close(void *agc_handle);
 
-#ifdef __cplusplus
-}
-#endif
-
 #endif // _ESP_AGC_H_

include/esp32s3/esp_mase.h

@@ -12,13 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License
 
-#ifndef _ESP_MASE_H_
-#define _ESP_MASE_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
 #define MASE_SAMPLE_RATE 16000        // Supports 16kHz only
 #define MASE_FRAME_SIZE 16            // Supports 16ms only
 #define MASE_MIC_DISTANCE 65          // According to physical design of mic-array
@@ -85,10 +78,4 @@ void mase_process(mase_handle_t st, int16_t *in, int16_t *dsp_out);
  * @return None
  *
  */
-void mase_destory(mase_handle_t st);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif
+void mase_destory(mase_handle_t st);

include/esp32s3/esp_vad.h

@@ -47,15 +47,11 @@ typedef void* vad_handle_t;
  *
  * @param vad_mode          Sets the VAD operating mode.
  *
- * @param sample_rate_hz    The Sampling frequency (Hz) can be 32000, 16000, 8000, default: 16000.
- *
- * @param one_frame_ms      The length of the audio processing can be 10ms, 20ms, 30ms, default: 30.
- *
  * @return
  *         - NULL: Create failed
  *         - Others: The instance of VAD
  */
-vad_handle_t vad_create(vad_mode_t vad_mode, int sample_rate_hz, int one_frame_ms);
+vad_handle_t vad_create(vad_mode_t vad_mode);
 
 /**
  * @brief Feed samples of an audio stream to the VAD and check if there is someone speaking.
@@ -64,12 +60,16 @@ vad_handle_t vad_create(vad_mode_t vad_mode, int sample_rate_hz, int one_frame_m
  *
  * @param data      An array of 16-bit signed audio samples.
  *
+ * @param sample_rate_hz    The Sampling frequency (Hz) can be 32000, 16000, 8000, default: 16000.
+ *
+ * @param one_frame_ms      The length of the audio processing can be 10ms, 20ms, 30ms, default: 30.
+ *
  * @return
  *         - VAD_SILENCE if no voice
  *         - VAD_SPEECH  if voice is detected
  *
  */
-vad_state_t vad_process(vad_handle_t inst, int16_t *data);
+vad_state_t vad_process(vad_handle_t inst, int16_t *data, int sample_rate_hz, int one_frame_ms);
 
 /**
  * @brief Free the VAD instance

libversion

@@ -1 +1 @@
-c008766c5e30abbd2e418086a688de59359ff1df
+fd54094054a442b011659a0d32898b728c33ad1c

model/movemodel.py

@@ -88,4 +88,4 @@ if multinet_model != 'null':
 # os.system("cp %s %s" % (wakenet_model+'/_MODEL_INFO_', target_model))
 
 total_size = calculate_total_size(target_model)
-print("Recommended model partition size: ", str(int((total_size / 1024 + 500) / 4 ) * 4) + 'KB')
+print("Recommended model partition size: ", str(int((total_size / 1024 + 500) / 4 ) * 4) + 'KB')

model/wakenet_model/alexa8/_MODEL_INFO_

@@ -1 +1 @@
-wakeNet8_v5h8_alexa_5_0.57_0.60
+wakeNet8_v5_alexa_5_0.55_0.54

model/wakenet_model/xiaoaitongxue7q8/_MODEL_INFO_

@@ -1 +1 @@
-wakeNet7Q8_v2h8_xiaoaitongxue_5_0.975_0.985
+wakeNet7Q8_v1_xiaoaitongxue_5_0.97_0.90