Boxlite aec (#465)

* 允许espbox-lite打开aec
微信模式使用小字体

* 软件回声消除

* 支持双麦采集

---------

Co-authored-by: cqm <cqm@97kid.com>
This commit is contained in:
mtdxc
2025-04-10 17:06:15 +08:00
committed by GitHub
parent 18e44ec775
commit f1ad29be3b
6 changed files with 61 additions and 26 deletions

View File

@@ -247,7 +247,7 @@ config USE_AUDIO_PROCESSOR
config USE_REALTIME_CHAT
bool "启用可语音打断的实时对话模式(需要 AEC 支持)"
default n
depends on USE_AUDIO_PROCESSOR && (BOARD_TYPE_ESP_BOX_3 || BOARD_TYPE_ESP_BOX || BOARD_TYPE_LICHUANG_DEV || BOARD_TYPE_ESP32S3_KORVO2_V3)
depends on USE_AUDIO_PROCESSOR && (BOARD_TYPE_ESP_BOX_3 || BOARD_TYPE_ESP_BOX || BOARD_TYPE_ESP_BOX_LITE || BOARD_TYPE_LICHUANG_DEV || BOARD_TYPE_ESP32S3_KORVO2_V3)
help
需要 ESP32 S3 与 AEC 开启,因为性能不够,不建议和微信聊天界面风格同时开启

View File

@@ -28,7 +28,7 @@ void AudioProcessor::Initialize(AudioCodec* codec, bool realtime_chat) {
afe_config_t* afe_config = afe_config_init(input_format.c_str(), NULL, AFE_TYPE_VC, AFE_MODE_HIGH_PERF);
if (realtime_chat) {
afe_config->aec_init = true;
afe_config->aec_mode = AEC_MODE_VOIP_LOW_COST;
afe_config->aec_mode = AEC_MODE_VOIP_HIGH_PERF;
} else {
afe_config->aec_init = false;
}

View File

@@ -3,6 +3,7 @@
#include <esp_log.h>
#include <driver/i2c_master.h>
#include <driver/i2s_tdm.h>
#include <cstring>
static const char TAG[] = "BoxAudioCodecLite";
@@ -11,7 +12,10 @@ BoxAudioCodecLite::BoxAudioCodecLite(void* i2c_master_handle, int input_sample_r
gpio_num_t pa_pin, bool input_reference) {
duplex_ = true; // 是否双工
input_reference_ = input_reference; // 是否使用参考输入,实现回声消除
input_channels_ = input_reference_ ? 2 : 1; // 输入通道数
if (input_reference) {
ref_buffer_.resize(960 * 2);
}
input_channels_ = 2 + input_reference_; // 输入通道数
input_sample_rate_ = input_sample_rate;
output_sample_rate_ = output_sample_rate;
@@ -108,18 +112,7 @@ void BoxAudioCodecLite::CreateDuplexChannels(gpio_num_t mclk, gpio_num_t bclk, g
.ext_clk_freq_hz = 0,
.mclk_multiple = I2S_MCLK_MULTIPLE_256
},
.slot_cfg = {
.data_bit_width = I2S_DATA_BIT_WIDTH_16BIT,
.slot_bit_width = I2S_SLOT_BIT_WIDTH_AUTO,
.slot_mode = I2S_SLOT_MODE_STEREO,
.slot_mask = I2S_STD_SLOT_BOTH,
.ws_width = I2S_DATA_BIT_WIDTH_16BIT,
.ws_pol = false,
.bit_shift = true,
.left_align = true,
.big_endian = false,
.bit_order_lsb = false
},
.slot_cfg = I2S_STD_PHILIP_SLOT_DEFAULT_CONFIG(I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_MONO),
.gpio_cfg = {
.mclk = mclk,
.bclk = bclk,
@@ -187,13 +180,13 @@ void BoxAudioCodecLite::EnableInput(bool enable) {
if (enable) {
esp_codec_dev_sample_info_t fs = {
.bits_per_sample = 16,
.channel = 4,
.channel_mask = ESP_CODEC_DEV_MAKE_CHANNEL_MASK(0),
.sample_rate = (uint32_t)output_sample_rate_,
.channel = (uint8_t)(input_channels_ - input_reference_),
.channel_mask = 0,
.sample_rate = (uint32_t)input_sample_rate_,
.mclk_multiple = 0,
};
if (input_reference_) {
fs.channel_mask |= ESP_CODEC_DEV_MAKE_CHANNEL_MASK(1);
for (int i = 0;i < fs.channel; i++) {
fs.channel_mask |= ESP_CODEC_DEV_MAKE_CHANNEL_MASK(i);
}
ESP_ERROR_CHECK(esp_codec_dev_open(input_dev_, &fs));
// 麦克风增益解决收音太小的问题
@@ -227,7 +220,30 @@ void BoxAudioCodecLite::EnableOutput(bool enable) {
int BoxAudioCodecLite::Read(int16_t* dest, int samples) {
if (input_enabled_) {
ESP_ERROR_CHECK_WITHOUT_ABORT(esp_codec_dev_read(input_dev_, (void*)dest, samples * sizeof(int16_t)));
if (!input_reference_) {
ESP_ERROR_CHECK_WITHOUT_ABORT(esp_codec_dev_read(input_dev_, (void*)dest, samples * sizeof(int16_t)));
}
else {
int size = samples / input_channels_;
int channels = input_channels_ - input_reference_;
std::vector<int16_t> data(size * channels);
// read mic data
ESP_ERROR_CHECK_WITHOUT_ABORT(esp_codec_dev_read(input_dev_, (void*)data.data(), data.size() * sizeof(int16_t)));
int j = 0;
int i = 0;
while (i< samples) {
// mic data
for (int p = 0; p < channels; p++) {
dest[i++] = data[j++];
}
// ref data
dest[i++] = read_pos_ < write_pos_? ref_buffer_[read_pos_++] : 0;
}
if (read_pos_ == write_pos_) {
read_pos_ = write_pos_ = 0;
}
}
}
return samples;
}
@@ -235,6 +251,22 @@ int BoxAudioCodecLite::Read(int16_t* dest, int samples) {
int BoxAudioCodecLite::Write(const int16_t* data, int samples) {
if (output_enabled_) {
ESP_ERROR_CHECK_WITHOUT_ABORT(esp_codec_dev_write(output_dev_, (void*)data, samples * sizeof(int16_t)));
if (input_reference_) { // 板子不支持硬件回采,采用缓存播放缓冲来实现回声消除
if (write_pos_ - read_pos_ + samples > ref_buffer_.size()) {
assert(ref_buffer_.size() >= samples);
// 写溢出,只保留最近的数据
read_pos_ = write_pos_ + samples - ref_buffer_.size();
}
if (read_pos_) {
if (write_pos_ != read_pos_) {
memmove(ref_buffer_.data(), ref_buffer_.data() + read_pos_, (write_pos_ - read_pos_) * sizeof(int16_t));
}
write_pos_ -= read_pos_;
read_pos_ = 0;
}
memcpy(&ref_buffer_[write_pos_], data, samples * sizeof(int16_t));
write_pos_ += samples;
}
}
return samples;
}

View File

@@ -17,6 +17,10 @@ private:
esp_codec_dev_handle_t output_dev_ = nullptr;
esp_codec_dev_handle_t input_dev_ = nullptr;
// ref buffer used for aec
std::vector<int16_t> ref_buffer_;
int read_pos_ = 0;
int write_pos_ = 0;
void CreateDuplexChannels(gpio_num_t mclk, gpio_num_t bclk, gpio_num_t ws, gpio_num_t dout, gpio_num_t din);

View File

@@ -3,10 +3,10 @@
#include <driver/gpio.h>
#define AUDIO_INPUT_SAMPLE_RATE 24000
#define AUDIO_OUTPUT_SAMPLE_RATE 24000
#define AUDIO_INPUT_SAMPLE_RATE 16000
#define AUDIO_OUTPUT_SAMPLE_RATE 16000
#define AUDIO_INPUT_REFERENCE true
#define AUDIO_INPUT_REFERENCE CONFIG_USE_AUDIO_PROCESSOR
#define AUDIO_I2S_GPIO_MCLK GPIO_NUM_2
#define AUDIO_I2S_GPIO_WS GPIO_NUM_47

View File

@@ -185,11 +185,10 @@ private:
esp_lcd_panel_reset(panel);
esp_lcd_panel_init(panel);
esp_lcd_panel_invert_color(panel, false);
esp_lcd_panel_invert_color(panel, DISPLAY_BACKLIGHT_OUTPUT_INVERT);
esp_lcd_panel_swap_xy(panel, DISPLAY_SWAP_XY);
esp_lcd_panel_mirror(panel, DISPLAY_MIRROR_X, DISPLAY_MIRROR_Y);
esp_lcd_panel_disp_on_off(panel, true);
esp_lcd_panel_invert_color(panel, true);
display_ = new SpiLcdDisplay(panel_io, panel,
DISPLAY_WIDTH, DISPLAY_HEIGHT, DISPLAY_OFFSET_X, DISPLAY_OFFSET_Y, DISPLAY_MIRROR_X, DISPLAY_MIRROR_Y, DISPLAY_SWAP_XY,
{