audio_demo/aecm_sender.cpp

561 lines
20 KiB
C++
Raw Normal View History

2025-02-14 08:58:27 +08:00
#include <iostream>
#include <unistd.h>
#include <cmath>
#include "timing.h"
#include "log/logger.h"
#include "common.h"
#include <modules/audio_processing/include/audio_processing.h>
#include <modules/audio_processing/include/config.h>
#include "alsa_dev.h"
using namespace std;
using namespace toolkit;
#define MIX_INPUT_CHANNELS 2
#define MIX_INPUT_SAMPLES (10 * MIX_INPUT_SAMPLE_RATE/1000)
// #define MIX_INPUT_SAMPLES 1024
#define MIX_INPUT_SAMPLE_RATE 44100
struct audio_buf_t
{
uint8_t* data;
int index;
int size;
};
struct RtmpConfig {
char url[1024];
AVFormatContext *formatCtx;
AVStream *stream;
AVCodecContext *codecCtx;
SwrContext *swrCtx;
};
static SampleInfo kPcmSampleInfo;
//----------------------------------------------
#define MIN(a, b) ((a) < (b) ? (a) : (b))
#define MAX(a, b) ((a) > (b) ? (a) : (b))
// 最大/小音量db
#define MIN_DB (-10)
#define MAX_DB (60)
// 最大/小音量: 0: 静音; 100:最大音量
#define MUTE_VOLUME (0)
#define MAX_VOLUME (100)
static int vol_scaler_init(int *scaler, int mindb, int maxdb);
typedef struct VolumeCtlUnit
{
int scaler[MAX_VOLUME + 1]; // 音量表
int zeroDb; // 0db在scaler中的索引
// 自定义需要调节的音量
int micVolume;
VolumeCtlUnit() {
// 音量控制器初始化
zeroDb = vol_scaler_init(scaler, MIN_DB, MAX_DB);
micVolume = 100;
}
} volume_ctl_unit_t;
static volume_ctl_unit_t kVolCtrlUnit;
static int vol_scaler_init(int *scaler, int mindb, int maxdb)
{
double tabdb[MAX_VOLUME + 1];
double tabf [MAX_VOLUME + 1];
int z, i;
for (i = 0; i < (MAX_VOLUME + 1); i++) {
// (mindb, maxdb)平均分成(MAX_VOLUME + 1)份
tabdb[i] = mindb + (maxdb - mindb) * i / (MAX_VOLUME + 1);
// dB = 20 * log(A1 / A2)当A1A2相等时db为0
// 这里以(1 << 14)作为原始声音振幅,得到调节后的振幅(A1),将A1存入音量表中
tabf [i] = pow(10.0, tabdb[i] / 20.0);
scaler[i] = (int)((1 << 14) * tabf[i]); // Q14 fix point
}
z = -mindb * (MAX_VOLUME + 1) / (maxdb - mindb);
z = MAX(z, 0 );
z = MIN(z, MAX_VOLUME);
scaler[0] = 0; // 音量表中0标识静音
scaler[z] = (1 << 14);// (mindb, maxdb)的中间值作为0db即不做增益处理
return z;
}
static void vol_scaler_run(int16_t *buf, int n, int volume)
{
/* 简易版
while (n--) {
*buf = (*buf) * multiplier / 100.0;
*buf = std::max((int)*buf, -0x7fff);
*buf = std::min((int)*buf, 0x7fff);
buf++;
}
*/
int multiplier = kVolCtrlUnit.scaler[volume];
if (multiplier > (1 << 14)) {
int32_t v;
while (n--) {
v = ((int32_t)*buf * multiplier) >> 14;
v = MAX(v,-0x7fff);
v = MIN(v, 0x7fff);
*buf++ = (int16_t)v;
}
} else if (multiplier < (1 << 14)) {
while (n--) {
*buf = ((int32_t)*buf * multiplier) >> 14;
buf++;
}
}
}
//----------------------------------------------
webrtc::AudioProcessing::Config webtcConfigInit()
{
webrtc::AudioProcessing::Config apmConfig;
apmConfig.pipeline.maximum_internal_processing_rate = MIX_INPUT_SAMPLE_RATE;
apmConfig.pipeline.multi_channel_capture = true;
apmConfig.pipeline.multi_channel_render = true;
//PreAmplifier
apmConfig.pre_amplifier.enabled = false;
apmConfig.pre_amplifier.fixed_gain_factor = 0.7f;
//HighPassFilter
apmConfig.high_pass_filter.enabled = false;
apmConfig.high_pass_filter.apply_in_full_band = false;
//EchoCanceller
apmConfig.echo_canceller.enabled = false;
apmConfig.echo_canceller.mobile_mode = false;
apmConfig.echo_canceller.export_linear_aec_output = false;
apmConfig.echo_canceller.enforce_high_pass_filtering = true;
//NoiseSuppression
apmConfig.noise_suppression.enabled = true;
apmConfig.noise_suppression.level = webrtc::AudioProcessing::Config::NoiseSuppression::kHigh;
apmConfig.noise_suppression.analyze_linear_aec_output_when_available = false;
//TransientSuppression
apmConfig.transient_suppression.enabled = false;
//VoiceDetection
apmConfig.voice_detection.enabled = true;
//GainController1
apmConfig.gain_controller1.enabled = true;
// kAdaptiveAnalog 自适应模拟模式
// kAdaptiveDigital 自适应数字增益模式
// kFixedDigital 固定数字增益模式
apmConfig.gain_controller1.mode = webrtc::AudioProcessing::Config::GainController1::kAdaptiveDigital;
apmConfig.gain_controller1.target_level_dbfs = 3; // 目标音量
apmConfig.gain_controller1.compression_gain_db = 12; // 增益能力
apmConfig.gain_controller1.enable_limiter = true; // 压限器开关
apmConfig.gain_controller1.analog_level_minimum = 0;
apmConfig.gain_controller1.analog_level_maximum = 496;
apmConfig.gain_controller1.analog_gain_controller.enabled = true;
apmConfig.gain_controller1.analog_gain_controller.startup_min_volume = webrtc::kAgcStartupMinVolume;
apmConfig.gain_controller1.analog_gain_controller.clipped_level_min = webrtc::kClippedLevelMin;
apmConfig.gain_controller1.analog_gain_controller.enable_agc2_level_estimator = false;
apmConfig.gain_controller1.analog_gain_controller.enable_digital_adaptive = true;
//GainController2
apmConfig.gain_controller2.enabled = false;
apmConfig.gain_controller2.fixed_digital.gain_db = 0.f;
apmConfig.gain_controller2.adaptive_digital.enabled = false;
apmConfig.gain_controller2.adaptive_digital.vad_probability_attack = 1.f;
apmConfig.gain_controller2.adaptive_digital.level_estimator = webrtc::AudioProcessing::Config::GainController2::kRms;
apmConfig.gain_controller2.adaptive_digital.level_estimator_adjacent_speech_frames_threshold = 1;
apmConfig.gain_controller2.adaptive_digital.use_saturation_protector = true;
apmConfig.gain_controller2.adaptive_digital.initial_saturation_margin_db = 20.f;
apmConfig.gain_controller2.adaptive_digital.extra_saturation_margin_db = 2.f;
apmConfig.gain_controller2.adaptive_digital.gain_applier_adjacent_speech_frames_threshold = 1;
apmConfig.gain_controller2.adaptive_digital.max_gain_change_db_per_second = 3.f;
apmConfig.gain_controller2.adaptive_digital.max_output_noise_level_dbfs = -50.f;
//ResidualEchoDetector
apmConfig.residual_echo_detector.enabled = false;
//LevelEstimation
apmConfig.level_estimation.enabled = false;
return apmConfig;
}
void pushDestory(RtmpConfig *config) {
if (config->formatCtx) {
if (config->formatCtx->pb)
avio_close(config->formatCtx->pb);
avformat_free_context(config->formatCtx);
}
if (config->codecCtx) {
avcodec_close(config->codecCtx);
avcodec_free_context(&config->codecCtx);
}
if (config->swrCtx) {
swr_close(config->swrCtx);
swr_free(&config->swrCtx);
}
}
int pushInit(RtmpConfig *config, int channels, AVSampleFormat format, int sample_rate)
{
if (nullptr == strstr(config->url, "rtmp://")) {
PrintE("url error, url: %s\n", config->url);
return -1;
}
AVCodec *codec = nullptr;
AVCodecContext *codecCtx = nullptr;
AVFormatContext *afctx = nullptr;
AVCodecParameters *codecPar = nullptr;
SwrContext *swrCtx = nullptr;
AVStream *audio_st = nullptr;
AVDictionary *opts = nullptr;
int ret;
// 打开输出流
ret = avformat_alloc_output_context2(&afctx, nullptr, "flv", config->url);
if (ret < 0) {
PrintE("open output failed.\n");
goto fail;
}
if ( !(afctx->oformat->flags & AVFMT_NOFILE) ) {
ret = avio_open(&afctx->pb, config->url, AVIO_FLAG_WRITE);
if (ret < 0) {
PrintE("avio_open failed.\n");
goto fail;
}
}
// 创建音频流
audio_st = avformat_new_stream(afctx, codec);
if (!audio_st) {
PrintE("alloc new audio stream failed.\n");
goto fail;
}
// 设置编码参数
codecPar = afctx->streams[audio_st->index]->codecpar;
codecPar->codec_id = AV_CODEC_ID_AAC;
codecPar->codec_type = AVMEDIA_TYPE_AUDIO;
codecPar->codec_tag = 0;
codecPar->bit_rate = 128 * 1024;
codecPar->sample_rate = sample_rate;
codecPar->channel_layout = av_get_default_channel_layout(channels);
codecPar->channels = av_get_channel_layout_nb_channels(codecPar->channel_layout);
codecPar->format = AV_SAMPLE_FMT_FLTP;
// 编码器初始化
codec = avcodec_find_encoder(codecPar->codec_id);
if (!codec) {
PrintE("find codec aac failed.\n");
return -1;
}
codecCtx = avcodec_alloc_context3(codec);
if (!codecCtx) {
PrintE("alloc codec context failed.\n");
goto fail;
}
ret = avcodec_parameters_to_context(codecCtx, codecPar);
if (ret < 0) {
PrintE("copt codec params failed.\n");
goto fail;
}
// 禁用缓冲
av_dict_set(&opts, "fflags", "nobuffer", AV_DICT_MATCH_CASE);
// av_dict_set(&opts, "rtmp_live", "1", AV_DICT_MATCH_CASE);
// 打开编码器
ret = avcodec_open2(codecCtx, codec, &opts);
if (ret < 0) {
PrintE("open codec {} failed.\n", codec->id);
goto fail;
}
audio_st->codecpar->codec_tag = 0;
// 释放字典资源
av_dict_free(&opts);
// 打印输出流信息
av_dump_format(afctx, 0, config->url, 1);
// 重采样初始化
swrCtx = swr_alloc_set_opts(nullptr,
codecCtx->channel_layout,
codecCtx->sample_fmt,
codecCtx->sample_rate,
av_get_default_channel_layout(channels),
format,
sample_rate,
0, nullptr);
if (!swrCtx) {
PrintE("swr_alloc_set_opts failed.\n");
goto fail;
}
swr_init(swrCtx);
config->codecCtx = codecCtx;
config->formatCtx = afctx;
config->stream = audio_st;
config->swrCtx = swrCtx;
PrintI("rtmp push init ok.\n");
return 0;
fail:
if (afctx) {
if (afctx->pb)
avio_close(afctx->pb);
avformat_free_context(afctx);
}
if (codecCtx) {
avcodec_close(codecCtx);
avcodec_free_context(&codecCtx);
}
if (swrCtx) {
swr_close(swrCtx);
swr_free(&swrCtx);
}
return -1;
}
int main(int argc, char *argv[])
{
if (argc < 4) {
fprintf(stderr, "usage %s card_num left(0) or right(1) url \n", argv[0]);
return -1;
}
//初始化日志系统
Logger::Instance().add(std::make_shared<ConsoleChannel> ());
Logger::Instance().add(std::make_shared<FileChannel>());
Logger::Instance().setWriter(std::make_shared<AsyncLogWriter>());
// 初始化声卡设备
int card = atoi(argv[1]);
int type = atoi(argv[2]);
alsa::Config alsaConfig;
alsaConfig.period_time = MIX_INPUT_SAMPLES * 1000000 / MIX_INPUT_SAMPLE_RATE;
alsaConfig.buffer_time = 5 * alsaConfig.period_time;
alsaConfig.channels = MIX_INPUT_CHANNELS;
alsaConfig.format = SND_PCM_FORMAT_S16_LE;
alsaConfig.rate = MIX_INPUT_SAMPLE_RATE;
if (card < 0)
sprintf(alsaConfig.device, "default");
else
sprintf(alsaConfig.device, "plughw:%d", card);
alsa::AlsaDev usbCaptureDev;
if (usbCaptureDev.applyConfig(alsaConfig) < 0) {
PrintE("alsa config failed.\n");
return -1;
}
PrintI("alsa before init: %s\n", usbCaptureDev.configToString());
if (usbCaptureDev.init(SND_PCM_STREAM_CAPTURE) < 0) {
PrintE("alsa init failed.\n");
return -1;
}
PrintI("alsa init: %s\n", usbCaptureDev.configToString());
// webrtc初始化
webrtc::AudioProcessing *apm = webrtc::AudioProcessingBuilder().Create();
if (!apm) {
PrintI("create apm failed.\n");
return -1;
}
webrtc::AudioProcessing::Config config = webtcConfigInit();
apm->ApplyConfig(config);
apm->Initialize();
apm->set_stream_analog_level(408);
webrtc::StreamConfig capConfig;
capConfig.set_has_keyboard(false);
capConfig.set_num_channels(alsaConfig.channels);
capConfig.set_sample_rate_hz(alsaConfig.rate);
PrintI("webrtc params: {\n%s\n}\n", config.ToString().c_str());
int ret = 0;
uint8_t *capData = nullptr;
int buffer_size = usbCaptureDev.getFrames() * usbCaptureDev.getFrameSize();
capData = (uint8_t *)malloc(buffer_size);
assert(capData);
uint8_t *listenBuffer = nullptr;
size_t listenSize = buffer_size;
listenBuffer = (uint8_t*)malloc(listenSize);
assert(listenBuffer);
// 推流初始化
RtmpConfig rtmp;
AVRational av;
int64_t pts = 0;
AVPacket *pkt = av_packet_alloc();
memset(&rtmp, 0, sizeof(rtmp));
strcpy(rtmp.url, argv[3]);
if (pushInit(&rtmp, MIX_INPUT_CHANNELS, AV_SAMPLE_FMT_S16, MIX_INPUT_SAMPLE_RATE) < 0) {
return -1;
}
av.den = rtmp.codecCtx->sample_rate;
av.num = 1;
AVFrame *inputFrame = av_frame_alloc();
{
inputFrame->sample_rate = rtmp.codecCtx->sample_rate;
inputFrame->format = rtmp.codecCtx->sample_fmt;
inputFrame->channels = rtmp.codecCtx->channels;
inputFrame->nb_samples = 1024;
inputFrame->channel_layout = rtmp.codecCtx->channel_layout;
int size = av_samples_get_buffer_size(nullptr,
rtmp.codecCtx->channels, inputFrame->nb_samples, rtmp.codecCtx->sample_fmt, 1);
uint8_t *buffer = (uint8_t *)av_malloc(size);
avcodec_fill_audio_frame(inputFrame, inputFrame->channels, rtmp.codecCtx->sample_fmt,
(const uint8_t*)buffer, size, 1);
}
AVFrame *outputFrame = av_frame_alloc();
{
outputFrame->format = rtmp.codecCtx->sample_fmt;
outputFrame->channel_layout = rtmp.codecCtx->channel_layout;
outputFrame->sample_rate = rtmp.codecCtx->sample_rate;
outputFrame->nb_samples = rtmp.codecCtx->frame_size;
int output_bz = av_samples_get_buffer_size(NULL, rtmp.codecCtx->channels, rtmp.codecCtx->frame_size, rtmp.codecCtx->sample_fmt, 0);
uint8_t *samples_data = (uint8_t *)av_malloc(output_bz);
avcodec_fill_audio_frame(outputFrame, rtmp.codecCtx->channels, rtmp.codecCtx->sample_fmt, samples_data, output_bz, 0);
}
// 写入帧头
ret = avformat_write_header(rtmp.formatCtx, nullptr);
if (ret < 0) {
PrintE("avformat_write_header failed.\n");
return ret;
}
FILE *ns_fp = fopen("/root/ns_out.pcm", "wb");
FILE *input_fp = fopen("/root/input_out.pcm", "wb");
std::vector<audio_buf_t> ns_out_list;
int frames = 0;
PrintI("input frame samples=%d, output frame samples=%d\n", inputFrame->nb_samples, outputFrame->nb_samples);
PrintI("------------------ start ------------------\n");
while (true)
{
// 采集
size_t read_size = usbCaptureDev.read(capData, buffer_size);
// PrintI("alsa read %d\n", read_size);
if (read_size <= 0) {
msleep(1);
continue;
}
// 分流
if (type >= 0) {
uint8_t *ptr = capData;
for (int i = 0; i < usbCaptureDev.getFrames(); ++i) {
// int size = av_get_bytes_per_sample(AV_SAMPLE_FMT_S16);
int size = usbCaptureDev.getSampleSize();
if (type == 0) {
// 左声道 (2.4G)
memcpy(listenBuffer + i * 2 * size, ptr + i * 2 * size, size);
memcpy(listenBuffer + (i * 2 + 1) * size, ptr + i * 2 * size, size);
} else {
// 右声道 (LINE IN)
memcpy(listenBuffer + i * 2 * size, ptr + (i * 2 + 1) * size, size);
memcpy(listenBuffer + (i * 2 + 1) * size, ptr + (i * 2 + 1) * size, size);
}
}
} else {
memcpy(listenBuffer, capData, read_size * usbCaptureDev.getFrameSize());
}
// vol_scaler_run((int16_t *)listenBuffer, capData->GetSamples() * kPcmSampleInfo.channels, kVolCtrlUnit.micVolume);
// 降噪
apm->ProcessStream((int16_t *)listenBuffer, capConfig, capConfig, (int16_t *)listenBuffer);
{
uint8_t *buffer = (uint8_t *)malloc(listenSize);
memcpy(buffer, listenBuffer, listenSize);
audio_buf_t out;
out.data = buffer;
out.index = 0;
out.size = listenSize;
ns_out_list.emplace_back(out);
}
// 填充音频
if (frames <= 0) frames = inputFrame->nb_samples;
while (frames > 0 && ns_out_list.size() > 0)
{
auto nsData = ns_out_list.begin();
int needSize = frames * sizeof(int16_t) * inputFrame->channels;
int readSize = (nsData->size - nsData->index) >= needSize ? needSize : (nsData->size - nsData->index);
memcpy(inputFrame->data[0] + (inputFrame->nb_samples - frames)*sizeof(int16_t)*inputFrame->channels, nsData->data + nsData->index, readSize);
frames -= readSize/(sizeof(int16_t) * inputFrame->channels);
nsData->index += readSize;
if (nsData->index >= nsData->size) {
free(nsData->data);
ns_out_list.erase(ns_out_list.begin());
}
}
if (frames > 0) continue;
// if (input_fp) fwrite(inputFrame->data[0], 1, inputFrame->nb_samples * inputFrame->channels * sizeof(int16_t), input_fp);
// 重采样
{
const uint8_t** in = (const uint8_t**)inputFrame->data;
uint8_t **out = outputFrame->data;
int len2, out_data_size;
len2 = swr_convert(rtmp.swrCtx, out, outputFrame->nb_samples, in, inputFrame->nb_samples);
if (len2 < 0) {
printf("swr_convert failed. \n");
break;
}
out_data_size = len2 * rtmp.codecCtx->channels * av_get_bytes_per_sample(rtmp.codecCtx->sample_fmt);
// if (ns_fp) fwrite(outputFrame->data[0], 1, out_data_size, ns_fp);
}
// 推流到远端
if (pts > INT64_MAX) pts = 0;
outputFrame->pts = pts;
pts += av_rescale_q(outputFrame->nb_samples, av, rtmp.codecCtx->time_base);
ret = avcodec_send_frame(rtmp.codecCtx, outputFrame);
if (ret < 0) {
PrintE("avcodec_send_frame failed: %d\n", ret);
break;
}
while (ret >= 0) {
ret = avcodec_receive_packet(rtmp.codecCtx, pkt);
if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF) {
break;
} else if (ret < 0) {
fprintf(stderr, "Error during encoding\n");
break;
}
// 将数据包时间戳从编码器时间基转换到流时间基
pkt->stream_index = rtmp.stream->index;
av_packet_rescale_ts(pkt, rtmp.codecCtx->time_base, rtmp.stream->time_base);
pkt->duration = av_rescale_q(pkt->duration, rtmp.codecCtx->time_base, rtmp.stream->time_base);
// 写入数据包到输出媒体文件
ret = av_interleaved_write_frame(rtmp.formatCtx, pkt);
if (ret < 0) {
fprintf(stderr, "Error while writing audio frame\n");
break;
}
// 释放数据包
av_packet_unref(pkt);
}
}
// 写入帧尾
av_write_trailer(rtmp.formatCtx);
if (apm) {
delete apm;
apm = nullptr;
}
pushDestory(&rtmp);
av_packet_free(&pkt);
av_frame_free(&inputFrame);
av_frame_free(&outputFrame);
av_free(listenBuffer);
for (auto buffer: ns_out_list) {
free(buffer.data);
}
return 0;
}