Nagram/TMessagesProj/jni/libtgvoip3/OpusDecoder.cpp
2020-04-24 12:21:58 +03:00

326 lines
10 KiB
C++

//
// libtgvoip is free and unencumbered public domain software.
// For more information, see http://unlicense.org or the UNLICENSE file
// you should have received with this source code distribution.
//
#include "logging.h"
#include "OpusDecoder.h"
#include "audio/Resampler.h"
#include <algorithm>
#include <cassert>
#include <cmath>
#include <cstring>
#if defined HAVE_CONFIG_H || defined TGVOIP_USE_INSTALLED_OPUS
#include <opus/opus.h>
#else
#include "opus.h"
#endif
#include "VoIPController.h"
#define PACKET_SIZE (960 * 2)
using namespace tgvoip;
tgvoip::OpusDecoder::OpusDecoder(MediaStreamItf* dst, bool isAsync, bool needEC)
{
dst->SetCallback(OpusDecoder::Callback, this);
Initialize(isAsync, needEC);
}
void tgvoip::OpusDecoder::Initialize(bool isAsync, bool needEC)
{
m_async = isAsync;
if (m_async)
{
m_decodedQueue = new BlockingQueue<Buffer>(33);
m_semaphore = new Semaphore(32, 0);
}
else
{
m_decodedQueue = nullptr;
m_semaphore = nullptr;
}
m_dec = opus_decoder_create(48000, 1, nullptr);
if (needEC)
m_ecDec = opus_decoder_create(48000, 1, nullptr);
else
m_ecDec = nullptr;
m_buffer=(unsigned char *) malloc(8192);
m_lastDecoded = nullptr;
m_outputBufferSize = 0;
m_echoCanceller = nullptr;
m_frameDuration = 20;
m_consecutiveLostPackets = 0;
m_enableDTX = false;
m_silentPacketCount = 0;
m_levelMeter = nullptr;
m_nextLen = 0;
m_running = false;
m_remainingDataLen = 0;
m_processedBuffer = nullptr;
m_prevWasEC = false;
m_prevLastSample = 0;
}
tgvoip::OpusDecoder::~OpusDecoder()
{
opus_decoder_destroy(m_dec);
if (m_ecDec)
opus_decoder_destroy(m_ecDec);
std::free(m_buffer);
delete m_decodedQueue;
delete m_semaphore;
}
void tgvoip::OpusDecoder::SetEchoCanceller(EchoCanceller* canceller)
{
m_echoCanceller = canceller;
}
std::size_t tgvoip::OpusDecoder::Callback(std::uint8_t* data, std::size_t len, void* param)
{
return (reinterpret_cast<OpusDecoder*>(param)->HandleCallback(data, len));
}
std::size_t tgvoip::OpusDecoder::HandleCallback(std::uint8_t* data, std::size_t len)
{
if (m_async)
{
if (!m_running)
{
std::memset(data, 0, len);
return 0;
}
if (m_outputBufferSize == 0)
{
m_outputBufferSize = len;
int packetsNeeded;
if (len > PACKET_SIZE)
packetsNeeded = static_cast<int>(len) / PACKET_SIZE;
else
packetsNeeded = 1;
packetsNeeded *= 2;
m_semaphore->Release(packetsNeeded);
}
assert(m_outputBufferSize == len && "output buffer size is supposed to be the same throughout callbacks");
if (len == PACKET_SIZE)
{
Buffer lastDecoded = m_decodedQueue->GetBlocking();
if (lastDecoded.IsEmpty())
return 0;
std::memcpy(data, *lastDecoded, PACKET_SIZE);
m_semaphore->Release();
if (m_silentPacketCount > 0)
{
--m_silentPacketCount;
if (m_levelMeter)
m_levelMeter->Update(reinterpret_cast<std::int16_t*>(data), 0);
return 0;
}
if (m_echoCanceller)
{
m_echoCanceller->SpeakerOutCallback(data, PACKET_SIZE);
}
}
else
{
LOGE("Opus decoder buffer length != 960 samples");
std::abort();
}
}
else
{
if (m_remainingDataLen == 0 && m_silentPacketCount == 0)
{
int duration = DecodeNextFrame();
m_remainingDataLen = static_cast<std::size_t>(duration) / 20 * 960 * 2;
}
if (m_silentPacketCount > 0 || m_remainingDataLen == 0 || m_processedBuffer == nullptr)
{
if (m_silentPacketCount > 0)
m_silentPacketCount--;
std::memset(data, 0, 960 * 2);
if (m_levelMeter)
m_levelMeter->Update(reinterpret_cast<std::int16_t*>(data), 0);
return 0;
}
std::memcpy(data, m_processedBuffer, 960 * 2);
m_remainingDataLen -= 960 * 2;
if (m_remainingDataLen > 0)
{
std::memmove(m_processedBuffer, m_processedBuffer + 960 * 2, static_cast<std::size_t>(m_remainingDataLen));
}
}
if (m_levelMeter != nullptr)
m_levelMeter->Update(reinterpret_cast<std::int16_t*>(data), len / 2);
return len;
}
void tgvoip::OpusDecoder::Start()
{
if (!m_async)
return;
m_running = true;
m_thread = new Thread(std::bind(&tgvoip::OpusDecoder::RunThread, this));
m_thread->SetName("opus_decoder");
m_thread->SetMaxPriority();
m_thread->Start();
}
void tgvoip::OpusDecoder::Stop()
{
if (!m_running || !m_async)
return;
m_running = false;
m_semaphore->Release();
m_thread->Join();
delete m_thread;
}
void tgvoip::OpusDecoder::RunThread()
{
LOGI("decoder: packets per frame %d", m_packetsPerFrame);
while (m_running)
{
int playbackDuration = DecodeNextFrame();
for (int i = 0; i < playbackDuration / 20; ++i)
{
m_semaphore->Acquire();
if (!m_running)
{
LOGI("==== decoder exiting ====");
return;
}
try
{
Buffer buf = m_bufferPool.Get();
if (m_remainingDataLen > 0)
{
for (effects::AudioEffect*& effect : m_postProcEffects)
{
effect->Process(reinterpret_cast<std::int16_t*>(m_processedBuffer + (PACKET_SIZE * i)), 960);
}
buf.CopyFrom(m_processedBuffer + (PACKET_SIZE * i), 0, PACKET_SIZE);
}
else
{
std::memset(*buf, 0, PACKET_SIZE);
}
m_decodedQueue->Put(std::move(buf));
}
catch (const std::bad_alloc& exception)
{
LOGW("decoder: no buffers left!\nwhat():\n%s", exception.what());
}
}
}
}
int tgvoip::OpusDecoder::DecodeNextFrame()
{
int playbackDuration = 0;
bool isEC = false;
std::size_t len = m_jitterBuffer->HandleOutput(m_buffer, 8192, 0, true, playbackDuration, isEC);
bool fec = false;
if (len == 0)
{
fec = true;
len = m_jitterBuffer->HandleOutput(m_buffer, 8192, 0, false, playbackDuration, isEC);
}
int size;
if (len != 0)
{
size = opus_decode(isEC ? m_ecDec : m_dec, m_buffer, len, reinterpret_cast<opus_int16*>(m_decodeBuffer), m_packetsPerFrame * 960, fec ? 1 : 0);
m_consecutiveLostPackets = 0;
if (m_prevWasEC != isEC && size)
{
// It turns out the waveforms generated by the PLC feature are also great to help smooth out the
// otherwise audible transition between the frames from different decoders. Those are basically an extrapolation
// of the previous successfully decoded data -- which is exactly what we need here.
size = opus_decode(m_prevWasEC ? m_ecDec : m_dec, nullptr, 0, reinterpret_cast<opus_int16*>(m_nextBuffer), m_packetsPerFrame * 960, 0);
if (size != 0)
{
std::int16_t* plcSamples = reinterpret_cast<std::int16_t*>(m_nextBuffer);
std::int16_t* samples = reinterpret_cast<std::int16_t*>(m_decodeBuffer);
constexpr float coeffs[] = { 0.999802f, 0.995062f, 0.984031f, 0.966778f, 0.943413f, 0.914084f, 0.878975f, 0.838309f, 0.792344f, 0.741368f,
0.685706f, 0.625708f, 0.561754f, 0.494249f, 0.423619f, 0.350311f, 0.274788f, 0.197527f, 0.119018f, 0.039757f };
for (int i = 0; i < 20; ++i)
{
samples[i] = static_cast<std::int16_t>(std::round(plcSamples[i] * coeffs[i] + samples[i] * (1.f - coeffs[i])));
}
}
}
m_prevWasEC = isEC;
m_prevLastSample = m_decodeBuffer[size - 1];
}
else
{ // do packet loss concealment
m_consecutiveLostPackets++;
if (m_consecutiveLostPackets > 2 && m_enableDTX)
{
m_silentPacketCount += m_packetsPerFrame;
size = static_cast<int>(m_packetsPerFrame) * 960;
}
else
{
size = opus_decode(m_prevWasEC ? m_ecDec : m_dec, nullptr, 0, reinterpret_cast<opus_int16*>(m_decodeBuffer), m_packetsPerFrame * 960, 0);
}
}
if (size < 0)
LOGW("decoder: opus_decode error %d", size);
m_remainingDataLen = size;
if (playbackDuration == 80)
{
m_processedBuffer = m_buffer;
audio::Resampler::Rescale60To80(reinterpret_cast<std::int16_t*>(m_decodeBuffer),
reinterpret_cast<std::int16_t*>(m_processedBuffer));
}
else if (playbackDuration == 40)
{
m_processedBuffer = m_buffer;
audio::Resampler::Rescale60To40(reinterpret_cast<std::int16_t*>(m_decodeBuffer),
reinterpret_cast<std::int16_t*>(m_processedBuffer));
}
else
{
m_processedBuffer = m_decodeBuffer;
}
return playbackDuration;
}
void tgvoip::OpusDecoder::SetFrameDuration(std::uint32_t duration)
{
m_frameDuration = duration;
m_packetsPerFrame = m_frameDuration / 20;
}
void tgvoip::OpusDecoder::SetJitterBuffer(std::shared_ptr<JitterBuffer> jitterBuffer)
{
m_jitterBuffer = std::move(jitterBuffer);
}
void tgvoip::OpusDecoder::SetDTX(bool enable)
{
m_enableDTX = enable;
}
void tgvoip::OpusDecoder::SetLevelMeter(AudioLevelMeter* levelMeter)
{
m_levelMeter = levelMeter;
}
void tgvoip::OpusDecoder::AddAudioEffect(effects::AudioEffect* effect)
{
m_postProcEffects.emplace_back(effect);
}
void tgvoip::OpusDecoder::RemoveAudioEffect(effects::AudioEffect* effect)
{
auto it = std::find(m_postProcEffects.begin(), m_postProcEffects.end(), effect);
if (it != m_postProcEffects.end())
m_postProcEffects.erase(it);
}