[i2s_audio] Correct a microphone with a DC offset signal (#8751)

esphome/components/i2s_audio/microphone/__init__.py

@@ -30,6 +30,7 @@ DEPENDENCIES = ["i2s_audio"]
 
 CONF_ADC_PIN = "adc_pin"
 CONF_ADC_TYPE = "adc_type"
+CONF_CORRECT_DC_OFFSET = "correct_dc_offset"
 CONF_PDM = "pdm"
 
 I2SAudioMicrophone = i2s_audio_ns.class_(
@@ -88,10 +89,13 @@ BASE_SCHEMA = microphone.MICROPHONE_SCHEMA.extend(
         default_sample_rate=16000,
         default_channel=CONF_RIGHT,
         default_bits_per_sample="32bit",
+    ).extend(
+        {
+            cv.Optional(CONF_CORRECT_DC_OFFSET, default=False): cv.boolean,
+        }
     )
 ).extend(cv.COMPONENT_SCHEMA)
 
-
 CONFIG_SCHEMA = cv.All(
     cv.typed_schema(
         {
@@ -140,3 +144,5 @@ async def to_code(config):
     else:
         cg.add(var.set_din_pin(config[CONF_I2S_DIN_PIN]))
         cg.add(var.set_pdm(config[CONF_PDM]))
+
+    cg.add(var.set_correct_dc_offset(config[CONF_CORRECT_DC_OFFSET]))

esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp

@@ -12,6 +12,8 @@
 #include "esphome/core/hal.h"
 #include "esphome/core/log.h"
 
+#include "esphome/components/audio/audio.h"
+
 namespace esphome {
 namespace i2s_audio {
 
@@ -22,6 +24,9 @@ static const uint32_t READ_DURATION_MS = 16;
 static const size_t TASK_STACK_SIZE = 4096;
 static const ssize_t TASK_PRIORITY = 23;
 
+// Use an exponential moving average to correct a DC offset with weight factor 1/1000
+static const int32_t DC_OFFSET_MOVING_AVERAGE_COEFFICIENT_DENOMINATOR = 1000;
+
 static const char *const TAG = "i2s_audio.microphone";
 
 enum MicrophoneEventGroupBits : uint32_t {
@@ -70,21 +75,11 @@ void I2SAudioMicrophone::setup() {
     this->mark_failed();
     return;
   }
+
+  this->configure_stream_settings_();
 }
 
-void I2SAudioMicrophone::start() {
+void I2SAudioMicrophone::configure_stream_settings_() {
-  if (this->is_failed())
-    return;
-
-  xSemaphoreTake(this->active_listeners_semaphore_, 0);
-}
-
-bool I2SAudioMicrophone::start_driver_() {
-  if (!this->parent_->try_lock()) {
-    return false;  // Waiting for another i2s to return lock
-  }
-  esp_err_t err;
-
   uint8_t channel_count = 1;
 #ifdef USE_I2S_LEGACY
   uint8_t bits_per_sample = this->bits_per_sample_;
@@ -93,10 +88,10 @@ bool I2SAudioMicrophone::start_driver_() {
     channel_count = 2;
   }
 #else
-  if (this->slot_bit_width_ == I2S_SLOT_BIT_WIDTH_AUTO) {
+  uint8_t bits_per_sample = 16;
-    this->slot_bit_width_ = I2S_SLOT_BIT_WIDTH_16BIT;
+  if (this->slot_bit_width_ != I2S_SLOT_BIT_WIDTH_AUTO) {
+    bits_per_sample = this->slot_bit_width_;
   }
-  uint8_t bits_per_sample = this->slot_bit_width_;
 
   if (this->slot_mode_ == I2S_SLOT_MODE_STEREO) {
     channel_count = 2;
@@ -114,6 +109,26 @@ bool I2SAudioMicrophone::start_driver_() {
   }
 #endif
 
+  if (this->pdm_) {
+    bits_per_sample = 16;  // PDM mics are always 16 bits per sample
+  }
+
+  this->audio_stream_info_ = audio::AudioStreamInfo(bits_per_sample, channel_count, this->sample_rate_);
+}
+
+void I2SAudioMicrophone::start() {
+  if (this->is_failed())
+    return;
+
+  xSemaphoreTake(this->active_listeners_semaphore_, 0);
+}
+
+bool I2SAudioMicrophone::start_driver_() {
+  if (!this->parent_->try_lock()) {
+    return false;  // Waiting for another i2s to return lock
+  }
+  esp_err_t err;
+
 #ifdef USE_I2S_LEGACY
   i2s_driver_config_t config = {
       .mode = (i2s_mode_t) (this->i2s_mode_ | I2S_MODE_RX),
@@ -202,8 +217,6 @@ bool I2SAudioMicrophone::start_driver_() {
   i2s_std_gpio_config_t pin_config = this->parent_->get_pin_config();
 #if SOC_I2S_SUPPORTS_PDM_RX
   if (this->pdm_) {
-    bits_per_sample = 16;  // PDM mics are always 16 bits per sample with the IDF 5 driver
-
     i2s_pdm_rx_clk_config_t clk_cfg = {
         .sample_rate_hz = this->sample_rate_,
         .clk_src = clk_src,
@@ -277,10 +290,8 @@ bool I2SAudioMicrophone::start_driver_() {
   }
 #endif
 
-  this->audio_stream_info_ = audio::AudioStreamInfo(bits_per_sample, channel_count, this->sample_rate_);
-
   this->status_clear_error();
-
+  this->configure_stream_settings_();  // redetermine the settings in case some settings were changed after compilation
   return true;
 }
 
@@ -361,9 +372,12 @@ void I2SAudioMicrophone::mic_task(void *params) {
         samples.resize(bytes_to_read);
         size_t bytes_read = this_microphone->read_(samples.data(), bytes_to_read, 2 * pdMS_TO_TICKS(READ_DURATION_MS));
         samples.resize(bytes_read);
+        if (this_microphone->correct_dc_offset_) {
+          this_microphone->fix_dc_offset_(samples);
+        }
         this_microphone->data_callbacks_.call(samples);
       } else {
-        delay(READ_DURATION_MS);
+        vTaskDelay(pdMS_TO_TICKS(READ_DURATION_MS));
       }
     }
   }
@@ -373,11 +387,34 @@ void I2SAudioMicrophone::mic_task(void *params) {
 
   xEventGroupSetBits(this_microphone->event_group_, MicrophoneEventGroupBits::TASK_STOPPED);
   while (true) {
-    // Continuously delay until the loop method delete the task
+    // Continuously delay until the loop method deletes the task
-    delay(10);
+    vTaskDelay(pdMS_TO_TICKS(10));
   }
 }
 
+void I2SAudioMicrophone::fix_dc_offset_(std::vector<uint8_t> &data) {
+  const size_t bytes_per_sample = this->audio_stream_info_.samples_to_bytes(1);
+  const uint32_t total_samples = this->audio_stream_info_.bytes_to_samples(data.size());
+
+  if (total_samples == 0) {
+    return;
+  }
+
+  int64_t offset_accumulator = 0;
+  for (uint32_t sample_index = 0; sample_index < total_samples; ++sample_index) {
+    const uint32_t byte_index = sample_index * bytes_per_sample;
+    int32_t sample = audio::unpack_audio_sample_to_q31(&data[byte_index], bytes_per_sample);
+    offset_accumulator += sample;
+    sample -= this->dc_offset_;
+    audio::pack_q31_as_audio_sample(sample, &data[byte_index], bytes_per_sample);
+  }
+
+  const int32_t new_offset = offset_accumulator / total_samples;
+  this->dc_offset_ = new_offset / DC_OFFSET_MOVING_AVERAGE_COEFFICIENT_DENOMINATOR +
+                     (DC_OFFSET_MOVING_AVERAGE_COEFFICIENT_DENOMINATOR - 1) * this->dc_offset_ /
+                         DC_OFFSET_MOVING_AVERAGE_COEFFICIENT_DENOMINATOR;
+}
+
 size_t I2SAudioMicrophone::read_(uint8_t *buf, size_t len, TickType_t ticks_to_wait) {
   size_t bytes_read = 0;
 #ifdef USE_I2S_LEGACY

esphome/components/i2s_audio/microphone/i2s_audio_microphone.h

@@ -7,8 +7,10 @@
 #include "esphome/components/microphone/microphone.h"
 #include "esphome/core/component.h"
 
+#include <freertos/FreeRTOS.h>
 #include <freertos/event_groups.h>
 #include <freertos/semphr.h>
+#include <freertos/task.h>
 
 namespace esphome {
 namespace i2s_audio {
@@ -20,6 +22,9 @@ class I2SAudioMicrophone : public I2SAudioIn, public microphone::Microphone, pub
   void stop() override;
 
   void loop() override;
+
+  void set_correct_dc_offset(bool correct_dc_offset) { this->correct_dc_offset_ = correct_dc_offset; }
+
 #ifdef USE_I2S_LEGACY
   void set_din_pin(int8_t pin) { this->din_pin_ = pin; }
 #else
@@ -41,8 +46,16 @@ class I2SAudioMicrophone : public I2SAudioIn, public microphone::Microphone, pub
   bool start_driver_();
   void stop_driver_();
 
+  /// @brief Attempts to correct a microphone DC offset; e.g., a microphones silent level is offset from 0. Applies a
+  /// correction offset that is updated using an exponential moving average for all samples away from 0.
+  /// @param data
+  void fix_dc_offset_(std::vector<uint8_t> &data);
+
   size_t read_(uint8_t *buf, size_t len, TickType_t ticks_to_wait);
 
+  /// @brief Sets the Microphone ``audio_stream_info_`` member variable to the configured I2S settings.
+  void configure_stream_settings_();
+
   static void mic_task(void *params);
 
   SemaphoreHandle_t active_listeners_semaphore_{nullptr};
@@ -61,6 +74,9 @@ class I2SAudioMicrophone : public I2SAudioIn, public microphone::Microphone, pub
   i2s_chan_handle_t rx_handle_;
 #endif
   bool pdm_{false};
+
+  bool correct_dc_offset_;
+  int32_t dc_offset_{0};
 };
 
 }  // namespace i2s_audio

tests/components/microphone/common.yaml

@@ -10,6 +10,7 @@ microphone:
     adc_type: external
     pdm: false
     mclk_multiple: 384
+    correct_dc_offset: true
     on_data:
       - if:
           condition: