fix(sr): Fix ESP-SR lib and example for v2.x

Author: me-no-dev

boards.txt

@@ -502,6 +502,10 @@ esp32p4.menu.PartitionScheme.app5M_little24M_32MB.upload.maximum_size=4718592
 esp32p4.menu.PartitionScheme.app13M_data7M_32MB=32M Flash (13MB APP/6.75MB SPIFFS)
 esp32p4.menu.PartitionScheme.app13M_data7M_32MB.build.partitions=default_32MB
 esp32p4.menu.PartitionScheme.app13M_data7M_32MB.upload.maximum_size=13107200
+esp32p4.menu.PartitionScheme.esp_sr_16=ESP SR 16M (3MB APP/7MB SPIFFS/2.9MB MODEL)
+esp32p4.menu.PartitionScheme.esp_sr_16.upload.maximum_size=3145728
+esp32p4.menu.PartitionScheme.esp_sr_16.upload.extra_flags=0xD10000 {build.path}/srmodels.bin
+esp32p4.menu.PartitionScheme.esp_sr_16.build.partitions=esp_sr_16
 esp32p4.menu.PartitionScheme.custom=Custom
 esp32p4.menu.PartitionScheme.custom.build.partitions=
 esp32p4.menu.PartitionScheme.custom.upload.maximum_size=16777216

libraries/ESP_SR/examples/Basic/Basic.ino

@@ -9,6 +9,17 @@
 #define LIGHT_PIN 40
 #define FAN_PIN 41
 
+/**
+ * The input format:
+ * M to represent the microphone channel
+ * R to represent the playback reference channel
+ * N to represent an unknown or unused channel
+ *
+ * For example, input_format="MMNR" indicates that the input data consists of four channels,
+ * which are the microphone channel, the microphone channel, an unused channel, and the playback channel
+ */
+#define SR_INPUT_FORMAT "MM"
+
 I2SClass i2s;
 
 // Generated using the following command:
@@ -69,7 +80,7 @@ void setup() {
  i2s.begin(I2S_MODE_STD, 16000, I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_STEREO);
 
  ESP_SR.onEvent(onSrEvent);
- ESP_SR.begin(i2s, sr_commands, sizeof(sr_commands) / sizeof(sr_cmd_t), SR_CHANNELS_STEREO, SR_MODE_WAKEWORD);
+ ESP_SR.begin(i2s, sr_commands, sizeof(sr_commands) / sizeof(sr_cmd_t), SR_CHANNELS_STEREO, SR_MODE_WAKEWORD, SR_INPUT_FORMAT);
 }
 
 void loop() {}

libraries/ESP_SR/examples/Basic/ci.json

@@ -2,6 +2,9 @@
  "fqbn": {
  "esp32s3": [
  "espressif:esp32:esp32s3:USBMode=default,PartitionScheme=esp_sr_16,FlashSize=16M,FlashMode=dio"
+ ],
+ "esp32p4": [
+ "espressif:esp32:esp32p4:USBMode=default,PartitionScheme=esp_sr_16,FlashSize=16M,FlashMode=qio"
  ]
  },
  "requires": [
@@ -12,7 +15,6 @@
  "esp32c3": false,
  "esp32c6": false,
  "esp32h2": false,
- "esp32p4": false,
  "esp32s2": false,
  "esp32c5": false
  }

libraries/ESP_SR/src/ESP_SR.cpp

@@ -4,7 +4,7 @@
  * SPDX-License-Identifier: Unlicense OR CC0-1.0
  */
 #include "sdkconfig.h"
-#if (CONFIG_IDF_TARGET_ESP32S3 || CONFIG_IDF_TARGET_ESP32P4) && (CONFIG_USE_WAKENET || CONFIG_USE_MULTINET)
+#if (CONFIG_IDF_TARGET_ESP32S3 || CONFIG_IDF_TARGET_ESP32P4) && (CONFIG_MODEL_IN_FLASH || CONFIG_MODEL_IN_SDCARD)
 #include "ESP_SR.h"
 
 static esp_err_t on_sr_fill(void *arg, void *out, size_t len, size_t *bytes_read, uint32_t timeout_ms) {
@@ -25,9 +25,9 @@ void ESP_SR_Class::onEvent(sr_cb event_cb) {
  cb = event_cb;
 }
 
-bool ESP_SR_Class::begin(I2SClass &_i2s, const sr_cmd_t *sr_commands, size_t sr_commands_len, sr_channels_t rx_chan, sr_mode_t mode) {
+bool ESP_SR_Class::begin(I2SClass &_i2s, const sr_cmd_t *sr_commands, size_t sr_commands_len, sr_channels_t rx_chan, sr_mode_t mode, const char *input_format) {
  i2s = &_i2s;
- esp_err_t err = sr_start(on_sr_fill, this, rx_chan, mode, sr_commands, sr_commands_len, on_sr_event, this);
+ esp_err_t err = sr_start(on_sr_fill, this, rx_chan, mode, input_format, sr_commands, sr_commands_len, on_sr_event, this);
  return (err == ESP_OK);
 }
 

libraries/ESP_SR/src/ESP_SR.h

@@ -6,7 +6,7 @@
 
 #pragma once
 #include "sdkconfig.h"
-#if (CONFIG_IDF_TARGET_ESP32S3 || CONFIG_IDF_TARGET_ESP32P4) && (CONFIG_USE_WAKENET || CONFIG_USE_MULTINET)
+#if (CONFIG_IDF_TARGET_ESP32S3 || CONFIG_IDF_TARGET_ESP32P4) && (CONFIG_MODEL_IN_FLASH || CONFIG_MODEL_IN_SDCARD)
 
 #include "ESP_I2S.h"
 #include "esp32-hal-sr.h"
@@ -23,7 +23,16 @@ class ESP_SR_Class {
  ~ESP_SR_Class();
 
  void onEvent(sr_cb cb);
- bool begin(I2SClass &i2s, const sr_cmd_t *sr_commands, size_t sr_commands_len, sr_channels_t rx_chan = SR_CHANNELS_STEREO, sr_mode_t mode = SR_MODE_WAKEWORD);
+/**
+ * The input format:
+ * M to represent the microphone channel
+ * R to represent the playback reference channel
+ * N to represent an unknown or unused channel
+ *
+ * For example, input_format="MMNR" indicates that the input data consists of four channels,
+ * which are the microphone channel, the microphone channel, an unused channel, and the playback channel
+ */
+ bool begin(I2SClass &i2s, const sr_cmd_t *sr_commands, size_t sr_commands_len, sr_channels_t rx_chan = SR_CHANNELS_STEREO, sr_mode_t mode = SR_MODE_WAKEWORD, const char *input_format = "MN");
  bool end(void);
  bool setMode(sr_mode_t mode);
  bool pause(void);

libraries/ESP_SR/src/esp32-hal-sr.c

@@ -4,7 +4,7 @@
  * SPDX-License-Identifier: Unlicense OR CC0-1.0
  */
 #include "sdkconfig.h"
-#if (CONFIG_IDF_TARGET_ESP32S3 || CONFIG_IDF_TARGET_ESP32P4) && (CONFIG_USE_WAKENET || CONFIG_USE_MULTINET)
+#if (CONFIG_IDF_TARGET_ESP32S3 || CONFIG_IDF_TARGET_ESP32P4) && (CONFIG_MODEL_IN_FLASH || CONFIG_MODEL_IN_SDCARD)
 
 #if !defined(ARDUINO_PARTITION_esp_sr_32) && !defined(ARDUINO_PARTITION_esp_sr_16) && !defined(ARDUINO_PARTITION_esp_sr_8)
 #warning Compatible partition must be selected for ESP_SR to work
@@ -313,7 +313,7 @@ esp_err_t sr_set_mode(sr_mode_t mode) {
 }
 
 esp_err_t sr_start(
- sr_fill_cb fill_cb, void *fill_cb_arg, sr_channels_t rx_chan, sr_mode_t mode, const sr_cmd_t sr_commands[], size_t cmd_number, sr_event_cb cb, void *cb_arg
+ sr_fill_cb fill_cb, void *fill_cb_arg, sr_channels_t rx_chan, sr_mode_t mode, const char *input_format, const sr_cmd_t sr_commands[], size_t cmd_number, sr_event_cb cb, void *cb_arg
 ) {
  esp_err_t ret = ESP_OK;
  ESP_RETURN_ON_FALSE(NULL == g_sr_data, ESP_ERR_INVALID_STATE, "SR already running");
@@ -340,12 +340,11 @@ esp_err_t sr_start(
  models = esp_srmodel_init("model");
 
  // Load WakeWord Detection
- g_sr_data->afe_handle = (esp_afe_sr_iface_t *)&ESP_AFE_SR_HANDLE;
- afe_config_t afe_config = AFE_CONFIG_DEFAULT();
- afe_config.wakenet_model_name = esp_srmodel_filter(models, ESP_WN_PREFIX, "hiesp");
- afe_config.aec_init = false;
+ afe_config_t *afe_config = afe_config_init(input_format, models, AFE_TYPE_SR, AFE_MODE_LOW_COST);
+ g_sr_data->afe_handle = esp_afe_handle_from_config(afe_config);
  log_d("load wakenet '%s'", afe_config.wakenet_model_name);
- g_sr_data->afe_data = g_sr_data->afe_handle->create_from_config(&afe_config);
+ g_sr_data->afe_data = g_sr_data->afe_handle->create_from_config(afe_config);
+ afe_config_free(afe_config);
 
  // Load Custom Command Detection
  char *mn_name = esp_srmodel_filter(models, ESP_MN_PREFIX, ESP_MN_ENGLISH);

libraries/ESP_SR/src/esp32-hal-sr.h

@@ -6,7 +6,7 @@
 
 #pragma once
 #include "sdkconfig.h"
-#if (CONFIG_IDF_TARGET_ESP32S3 || CONFIG_IDF_TARGET_ESP32P4) && (CONFIG_USE_WAKENET || CONFIG_USE_MULTINET)
+#if (CONFIG_IDF_TARGET_ESP32S3 || CONFIG_IDF_TARGET_ESP32P4) && (CONFIG_MODEL_IN_FLASH || CONFIG_MODEL_IN_SDCARD)
 
 #include "driver/i2s_types.h"
 #include "esp_err.h"
@@ -49,7 +49,7 @@ typedef void (*sr_event_cb)(void *arg, sr_event_t event, int command_id, int phr
 typedef esp_err_t (*sr_fill_cb)(void *arg, void *out, size_t len, size_t *bytes_read, uint32_t timeout_ms);
 
 esp_err_t sr_start(
- sr_fill_cb fill_cb, void *fill_cb_arg, sr_channels_t rx_chan, sr_mode_t mode, const sr_cmd_t *sr_commands, size_t cmd_number, sr_event_cb cb, void *cb_arg
+ sr_fill_cb fill_cb, void *fill_cb_arg, sr_channels_t rx_chan, sr_mode_t mode, const char *input_format, const sr_cmd_t *sr_commands, size_t cmd_number, sr_event_cb cb, void *cb_arg
 );
 esp_err_t sr_stop(void);
 esp_err_t sr_pause(void);