Merge branch 'feat/threshold' into 'master'

feat: add more interface for afe to modify threshold of wakenet

See merge request speech-recognition-framework/esp-sr!167

include/esp32/esp_afe_sr_iface.h

@@ -131,14 +131,23 @@ typedef afe_fetch_result_t *(*esp_afe_sr_iface_op_fetch_with_delay_t)(esp_afe_sr
 typedef int (*esp_afe_sr_iface_op_reset_buffer_t)(esp_afe_sr_data_t *afe);
 
 /**
- * @brief Initial wakenet and wake words coefficient, or reset wakenet and wake words coefficient
- *        when wakenet has been initialized. It's only support wakenet 1 now.
- *
- * @param afe                The AFE_SR object to query
- * @param wakenet_word       The wakenet word, should be DEFAULT_WAKE_WORD or EXTRA_WAKE_WORD
+ * @brief Set wakenet detection threshold 
+ * 
+ * @param afe           The AFE_SR object to query
+ * @param index         The wakenet index, just support 1: wakenet1 or  2: wakenet2
+ * @param threshold     The wakenet detection threshold, the value is between 0.4 and 0.9999.
  * @return             -1: fail, 1: success
  */
-typedef int (*esp_afe_sr_iface_op_set_wakenet_t)(esp_afe_sr_data_t *afe, char *model_name);
+typedef int (*esp_afe_sr_iface_op_set_wakenet_threshold_t)(esp_afe_sr_data_t *afe, int index, float threshold);
+
+/**
+ * @brief Reset wakenet detection threshold to inital state
+ * 
+ * @param afe           The AFE_SR object to query
+ * @param index         The wakenet index, just support 1: wakenet1 or  2: wakenet2
+ * @return             -1: fail, 1: success
+ */
+typedef int (*esp_afe_sr_iface_op_reset_wakenet_threshold_t)(esp_afe_sr_data_t *afe, int index);
 
 /**
  * @brief Reset one function/module/algorithm.
@@ -195,7 +204,8 @@ typedef struct {
     esp_afe_sr_iface_op_get_channel_num_t get_feed_channel_num;
     esp_afe_sr_iface_op_get_channel_num_t get_fetch_channel_num;
     esp_afe_sr_iface_op_get_samp_rate_t get_samp_rate;
-    esp_afe_sr_iface_op_set_wakenet_t set_wakenet;
+    esp_afe_sr_iface_op_set_wakenet_threshold_t set_wakenet_threshold;
+    esp_afe_sr_iface_op_reset_wakenet_threshold_t reset_wakenet_threshold;
     esp_afe_sr_iface_op_disable_func_t disable_wakenet;
     esp_afe_sr_iface_op_enable_func_t enable_wakenet;
     esp_afe_sr_iface_op_disable_func_t disable_aec;

include/esp32/esp_wn_iface.h

@@ -110,12 +110,21 @@ typedef char* (*esp_wn_iface_op_get_word_name_t)(model_iface_data_t *model, int
  * @brief Set the detection threshold to manually abjust the probability 
  *
  * @param model The model object to query
- * @param det_treshold The threshold to trigger wake words, the range of det_threshold is 0.5~0.9999
+ * @param det_treshold The threshold to trigger wake words, the range of det_threshold is 0.4~0.9999
  * @param word_index The index of wake word
  * @return 0: setting failed, 1: setting success
  */
 typedef int (*esp_wn_iface_op_set_det_threshold_t)(model_iface_data_t *model, float det_threshold, int word_index);
 
+/**
+ * @brief Reset the threshold to its initial state  
+ *
+ * @param model The model object to query
+ * @return 0: setting failed, 1: setting success
+ */
+typedef int (*esp_wn_iface_op_reset_det_threshold_t)(model_iface_data_t *model);
+
+
 /**
  * @brief Get the wake word detection threshold of different modes
  *
@@ -200,6 +209,7 @@ typedef struct {
     esp_wn_iface_op_get_word_num_t get_word_num;
     esp_wn_iface_op_get_word_name_t get_word_name;
     esp_wn_iface_op_set_det_threshold_t set_det_threshold;
+    esp_wn_iface_op_reset_det_threshold_t reset_det_threshold;
     esp_wn_iface_op_get_det_threshold_t get_det_threshold;
     esp_wn_iface_op_get_triggered_channel_t  get_triggered_channel;
     esp_wn_iface_op_get_vol_gain_t get_vol_gain;

include/esp32p4/esp_afe_sr_iface.h

@@ -131,14 +131,23 @@ typedef afe_fetch_result_t *(*esp_afe_sr_iface_op_fetch_with_delay_t)(esp_afe_sr
 typedef int (*esp_afe_sr_iface_op_reset_buffer_t)(esp_afe_sr_data_t *afe);
 
 /**
- * @brief Initial wakenet and wake words coefficient, or reset wakenet and wake words coefficient
- *        when wakenet has been initialized. It's only support wakenet 1 now.
- *
- * @param afe                The AFE_SR object to query
- * @param wakenet_word       The wakenet word, should be DEFAULT_WAKE_WORD or EXTRA_WAKE_WORD
+ * @brief Set wakenet detection threshold 
+ * 
+ * @param afe           The AFE_SR object to query
+ * @param index         The wakenet index, just support 1: wakenet1 or  2: wakenet2
+ * @param threshold     The wakenet detection threshold, the value is between 0.4 and 0.9999.
  * @return             -1: fail, 1: success
  */
-typedef int (*esp_afe_sr_iface_op_set_wakenet_t)(esp_afe_sr_data_t *afe, char *model_name);
+typedef int (*esp_afe_sr_iface_op_set_wakenet_threshold_t)(esp_afe_sr_data_t *afe, int index, float threshold);
+
+/**
+ * @brief Reset wakenet detection threshold to inital state
+ * 
+ * @param afe           The AFE_SR object to query
+ * @param index         The wakenet index, just support 1: wakenet1 or  2: wakenet2
+ * @return             -1: fail, 1: success
+ */
+typedef int (*esp_afe_sr_iface_op_reset_wakenet_threshold_t)(esp_afe_sr_data_t *afe, int index);
 
 /**
  * @brief Reset one function/module/algorithm.
@@ -195,7 +204,8 @@ typedef struct {
     esp_afe_sr_iface_op_get_channel_num_t get_feed_channel_num;
     esp_afe_sr_iface_op_get_channel_num_t get_fetch_channel_num;
     esp_afe_sr_iface_op_get_samp_rate_t get_samp_rate;
-    esp_afe_sr_iface_op_set_wakenet_t set_wakenet;
+    esp_afe_sr_iface_op_set_wakenet_threshold_t set_wakenet_threshold;
+    esp_afe_sr_iface_op_reset_wakenet_threshold_t reset_wakenet_threshold;
     esp_afe_sr_iface_op_disable_func_t disable_wakenet;
     esp_afe_sr_iface_op_enable_func_t enable_wakenet;
     esp_afe_sr_iface_op_disable_func_t disable_aec;

include/esp32p4/esp_wn_iface.h

@@ -110,12 +110,21 @@ typedef char* (*esp_wn_iface_op_get_word_name_t)(model_iface_data_t *model, int
  * @brief Set the detection threshold to manually abjust the probability 
  *
  * @param model The model object to query
- * @param det_treshold The threshold to trigger wake words, the range of det_threshold is 0.5~0.9999
+ * @param det_treshold The threshold to trigger wake words, the range of det_threshold is 0.4~0.9999
  * @param word_index The index of wake word
  * @return 0: setting failed, 1: setting success
  */
 typedef int (*esp_wn_iface_op_set_det_threshold_t)(model_iface_data_t *model, float det_threshold, int word_index);
 
+/**
+ * @brief Reset the threshold to its initial state  
+ *
+ * @param model The model object to query
+ * @return 0: setting failed, 1: setting success
+ */
+typedef int (*esp_wn_iface_op_reset_det_threshold_t)(model_iface_data_t *model);
+
+
 /**
  * @brief Get the wake word detection threshold of different modes
  *
@@ -200,6 +209,7 @@ typedef struct {
     esp_wn_iface_op_get_word_num_t get_word_num;
     esp_wn_iface_op_get_word_name_t get_word_name;
     esp_wn_iface_op_set_det_threshold_t set_det_threshold;
+    esp_wn_iface_op_reset_det_threshold_t reset_det_threshold;
     esp_wn_iface_op_get_det_threshold_t get_det_threshold;
     esp_wn_iface_op_get_triggered_channel_t  get_triggered_channel;
     esp_wn_iface_op_get_vol_gain_t get_vol_gain;

include/esp32s3/esp_afe_sr_iface.h

@@ -131,14 +131,23 @@ typedef afe_fetch_result_t *(*esp_afe_sr_iface_op_fetch_with_delay_t)(esp_afe_sr
 typedef int (*esp_afe_sr_iface_op_reset_buffer_t)(esp_afe_sr_data_t *afe);
 
 /**
- * @brief Initial wakenet and wake words coefficient, or reset wakenet and wake words coefficient
- *        when wakenet has been initialized. It's only support wakenet 1 now.
- *
- * @param afe                The AFE_SR object to query
- * @param wakenet_word       The wakenet word, should be DEFAULT_WAKE_WORD or EXTRA_WAKE_WORD
+ * @brief Set wakenet detection threshold 
+ * 
+ * @param afe           The AFE_SR object to query
+ * @param index         The wakenet index, just support 1: wakenet1 or  2: wakenet2
+ * @param threshold     The wakenet detection threshold, the value is between 0.4 and 0.9999.
  * @return             -1: fail, 1: success
  */
-typedef int (*esp_afe_sr_iface_op_set_wakenet_t)(esp_afe_sr_data_t *afe, char *model_name);
+typedef int (*esp_afe_sr_iface_op_set_wakenet_threshold_t)(esp_afe_sr_data_t *afe, int index, float threshold);
+
+/**
+ * @brief Reset wakenet detection threshold to inital state
+ * 
+ * @param afe           The AFE_SR object to query
+ * @param index         The wakenet index, just support 1: wakenet1 or  2: wakenet2
+ * @return             -1: fail, 1: success
+ */
+typedef int (*esp_afe_sr_iface_op_reset_wakenet_threshold_t)(esp_afe_sr_data_t *afe, int index);
 
 /**
  * @brief Reset one function/module/algorithm.
@@ -195,7 +204,8 @@ typedef struct {
     esp_afe_sr_iface_op_get_channel_num_t get_feed_channel_num;
     esp_afe_sr_iface_op_get_channel_num_t get_fetch_channel_num;
     esp_afe_sr_iface_op_get_samp_rate_t get_samp_rate;
-    esp_afe_sr_iface_op_set_wakenet_t set_wakenet;
+    esp_afe_sr_iface_op_set_wakenet_threshold_t set_wakenet_threshold;
+    esp_afe_sr_iface_op_reset_wakenet_threshold_t reset_wakenet_threshold;
     esp_afe_sr_iface_op_disable_func_t disable_wakenet;
     esp_afe_sr_iface_op_enable_func_t enable_wakenet;
     esp_afe_sr_iface_op_disable_func_t disable_aec;

include/esp32s3/esp_wn_iface.h

@@ -110,12 +110,21 @@ typedef char* (*esp_wn_iface_op_get_word_name_t)(model_iface_data_t *model, int
  * @brief Set the detection threshold to manually abjust the probability 
  *
  * @param model The model object to query
- * @param det_treshold The threshold to trigger wake words, the range of det_threshold is 0.5~0.9999
+ * @param det_treshold The threshold to trigger wake words, the range of det_threshold is 0.4~0.9999
  * @param word_index The index of wake word
  * @return 0: setting failed, 1: setting success
  */
 typedef int (*esp_wn_iface_op_set_det_threshold_t)(model_iface_data_t *model, float det_threshold, int word_index);
 
+/**
+ * @brief Reset the threshold to its initial state  
+ *
+ * @param model The model object to query
+ * @return 0: setting failed, 1: setting success
+ */
+typedef int (*esp_wn_iface_op_reset_det_threshold_t)(model_iface_data_t *model);
+
+
 /**
  * @brief Get the wake word detection threshold of different modes
  *
@@ -200,6 +209,7 @@ typedef struct {
     esp_wn_iface_op_get_word_num_t get_word_num;
     esp_wn_iface_op_get_word_name_t get_word_name;
     esp_wn_iface_op_set_det_threshold_t set_det_threshold;
+    esp_wn_iface_op_reset_det_threshold_t reset_det_threshold;
     esp_wn_iface_op_get_det_threshold_t get_det_threshold;
     esp_wn_iface_op_get_triggered_channel_t  get_triggered_channel;
     esp_wn_iface_op_get_vol_gain_t get_vol_gain;

test_apps/esp-sr/main/test_afe.cpp

@@ -60,6 +60,8 @@ void test_afe_by_config(afe_config_t *afe_config, int frame_num, int *memory, fl
         // run afe feed
         feed_chunksize = afe_handle->get_feed_chunksize(afe_data);
         int feed_nch = afe_handle->get_feed_channel_num(afe_data);
+        afe_handle->set_wakenet_threshold(afe_data, 1, 0.8);
+        afe_handle->reset_wakenet_threshold(afe_data, 1);
 
         int16_t *feed_buff = (int16_t *)malloc(feed_chunksize * sizeof(int16_t) * feed_nch);
         start = esp_timer_get_time();

test_apps/esp-sr/main/test_wakenet.cpp

@@ -92,6 +92,8 @@ TEST_CASE("wakenet detect API & cpu loading", "[wn]")
     int16_t *buffer = (int16_t *) malloc(audio_chunksize);
     int chunks = 0;
     int detected = 0;
+    wakenet->set_det_threshold(model_data, 0.8, 1);
+    wakenet->reset_det_threshold(model_data);
     struct timeval tv_start, tv_end;
     gettimeofday(&tv_start, NULL);
     unsigned char* data = NULL;