2017-02-02 17:24:40 +01:00
|
|
|
//
|
|
|
|
// 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 "OpusDecoder.h"
|
2017-05-21 16:50:23 +02:00
|
|
|
#include "audio/Resampler.h"
|
2017-02-02 17:24:40 +01:00
|
|
|
#include "logging.h"
|
|
|
|
#include <assert.h>
|
2018-06-04 23:28:38 +02:00
|
|
|
#include <math.h>
|
2017-09-07 07:39:33 +02:00
|
|
|
#include <algorithm>
|
2019-11-18 09:44:18 +01:00
|
|
|
#if defined HAVE_CONFIG_H || defined TGVOIP_USE_INSTALLED_OPUS
|
2018-07-17 18:48:21 +02:00
|
|
|
#include <opus/opus.h>
|
|
|
|
#else
|
2020-01-18 15:36:59 +01:00
|
|
|
#include "opus.h"
|
2018-07-17 18:48:21 +02:00
|
|
|
#endif
|
2017-02-02 17:24:40 +01:00
|
|
|
|
2018-05-15 20:23:46 +02:00
|
|
|
#include "VoIPController.h"
|
|
|
|
|
2017-02-02 17:24:40 +01:00
|
|
|
#define PACKET_SIZE (960*2)
|
|
|
|
|
2017-04-28 13:17:56 +02:00
|
|
|
using namespace tgvoip;
|
|
|
|
|
2018-06-04 21:37:43 +02:00
|
|
|
tgvoip::OpusDecoder::OpusDecoder(const std::shared_ptr<MediaStreamItf>& dst, bool isAsync, bool needEC){
|
|
|
|
dst->SetCallback(OpusDecoder::Callback, this);
|
|
|
|
Initialize(isAsync, needEC);
|
|
|
|
}
|
|
|
|
|
|
|
|
tgvoip::OpusDecoder::OpusDecoder(const std::unique_ptr<MediaStreamItf>& dst, bool isAsync, bool needEC){
|
2017-04-28 13:17:56 +02:00
|
|
|
dst->SetCallback(OpusDecoder::Callback, this);
|
2018-06-04 21:37:43 +02:00
|
|
|
Initialize(isAsync, needEC);
|
|
|
|
}
|
|
|
|
|
2018-07-17 18:48:21 +02:00
|
|
|
tgvoip::OpusDecoder::OpusDecoder(MediaStreamItf* dst, bool isAsync, bool needEC){
|
|
|
|
dst->SetCallback(OpusDecoder::Callback, this);
|
|
|
|
Initialize(isAsync, needEC);
|
|
|
|
}
|
|
|
|
|
2018-06-04 21:37:43 +02:00
|
|
|
void tgvoip::OpusDecoder::Initialize(bool isAsync, bool needEC){
|
|
|
|
async=isAsync;
|
2018-05-15 20:23:46 +02:00
|
|
|
if(async){
|
|
|
|
decodedQueue=new BlockingQueue<unsigned char*>(33);
|
|
|
|
bufferPool=new BufferPool(PACKET_SIZE, 32);
|
|
|
|
semaphore=new Semaphore(32, 0);
|
|
|
|
}else{
|
|
|
|
decodedQueue=NULL;
|
|
|
|
bufferPool=NULL;
|
|
|
|
semaphore=NULL;
|
|
|
|
}
|
2017-02-02 17:24:40 +01:00
|
|
|
dec=opus_decoder_create(48000, 1, NULL);
|
2018-06-04 21:37:43 +02:00
|
|
|
if(needEC)
|
|
|
|
ecDec=opus_decoder_create(48000, 1, NULL);
|
|
|
|
else
|
|
|
|
ecDec=NULL;
|
2017-05-21 16:50:23 +02:00
|
|
|
buffer=(unsigned char *) malloc(8192);
|
2017-02-02 17:24:40 +01:00
|
|
|
lastDecoded=NULL;
|
|
|
|
outputBufferSize=0;
|
|
|
|
echoCanceller=NULL;
|
|
|
|
frameDuration=20;
|
2018-05-15 20:23:46 +02:00
|
|
|
consecutiveLostPackets=0;
|
|
|
|
enableDTX=false;
|
|
|
|
silentPacketCount=0;
|
|
|
|
levelMeter=NULL;
|
|
|
|
nextLen=0;
|
|
|
|
running=false;
|
|
|
|
remainingDataLen=0;
|
|
|
|
processedBuffer=NULL;
|
2018-06-04 21:37:43 +02:00
|
|
|
prevWasEC=false;
|
|
|
|
prevLastSample=0;
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
|
2017-04-28 13:17:56 +02:00
|
|
|
tgvoip::OpusDecoder::~OpusDecoder(){
|
2017-02-02 17:24:40 +01:00
|
|
|
opus_decoder_destroy(dec);
|
2018-06-04 21:37:43 +02:00
|
|
|
if(ecDec)
|
|
|
|
opus_decoder_destroy(ecDec);
|
2017-02-02 17:24:40 +01:00
|
|
|
free(buffer);
|
2018-05-15 20:23:46 +02:00
|
|
|
if(bufferPool)
|
|
|
|
delete bufferPool;
|
|
|
|
if(decodedQueue)
|
|
|
|
delete decodedQueue;
|
|
|
|
if(semaphore)
|
|
|
|
delete semaphore;
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2017-04-28 13:17:56 +02:00
|
|
|
void tgvoip::OpusDecoder::SetEchoCanceller(EchoCanceller* canceller){
|
2017-02-02 17:24:40 +01:00
|
|
|
echoCanceller=canceller;
|
|
|
|
}
|
|
|
|
|
2017-04-28 13:17:56 +02:00
|
|
|
size_t tgvoip::OpusDecoder::Callback(unsigned char *data, size_t len, void *param){
|
2018-05-15 20:23:46 +02:00
|
|
|
return ((OpusDecoder*)param)->HandleCallback(data, len);
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
|
2018-05-15 20:23:46 +02:00
|
|
|
size_t tgvoip::OpusDecoder::HandleCallback(unsigned char *data, size_t len){
|
|
|
|
if(async){
|
|
|
|
if(!running){
|
|
|
|
memset(data, 0, len);
|
|
|
|
return 0;
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
2018-05-15 20:23:46 +02:00
|
|
|
if(outputBufferSize==0){
|
|
|
|
outputBufferSize=len;
|
|
|
|
int packetsNeeded;
|
|
|
|
if(len>PACKET_SIZE)
|
|
|
|
packetsNeeded=len/PACKET_SIZE;
|
2017-02-02 17:24:40 +01:00
|
|
|
else
|
2018-05-15 20:23:46 +02:00
|
|
|
packetsNeeded=1;
|
|
|
|
packetsNeeded*=2;
|
|
|
|
semaphore->Release(packetsNeeded);
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
2018-05-15 20:23:46 +02:00
|
|
|
assert(outputBufferSize==len && "output buffer size is supposed to be the same throughout callbacks");
|
2017-02-02 17:24:40 +01:00
|
|
|
if(len==PACKET_SIZE){
|
2018-05-15 20:23:46 +02:00
|
|
|
lastDecoded=(unsigned char *) decodedQueue->GetBlocking();
|
|
|
|
if(!lastDecoded)
|
|
|
|
return 0;
|
|
|
|
memcpy(data, lastDecoded, PACKET_SIZE);
|
|
|
|
bufferPool->Reuse(lastDecoded);
|
|
|
|
semaphore->Release();
|
|
|
|
if(silentPacketCount>0){
|
|
|
|
silentPacketCount--;
|
|
|
|
if(levelMeter)
|
|
|
|
levelMeter->Update(reinterpret_cast<int16_t *>(data), 0);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
if(echoCanceller){
|
|
|
|
echoCanceller->SpeakerOutCallback(data, PACKET_SIZE);
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
2018-05-15 20:23:46 +02:00
|
|
|
}else{
|
|
|
|
LOGE("Opus decoder buffer length != 960 samples");
|
|
|
|
abort();
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
}else{
|
2018-05-15 20:23:46 +02:00
|
|
|
if(remainingDataLen==0 && silentPacketCount==0){
|
|
|
|
int duration=DecodeNextFrame();
|
|
|
|
remainingDataLen=(size_t) (duration/20*960*2);
|
|
|
|
}
|
|
|
|
if(silentPacketCount>0 || remainingDataLen==0 || !processedBuffer){
|
|
|
|
if(silentPacketCount>0)
|
|
|
|
silentPacketCount--;
|
|
|
|
memset(data, 0, 960*2);
|
|
|
|
if(levelMeter)
|
|
|
|
levelMeter->Update(reinterpret_cast<int16_t *>(data), 0);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
memcpy(data, processedBuffer, 960*2);
|
|
|
|
remainingDataLen-=960*2;
|
|
|
|
if(remainingDataLen>0){
|
|
|
|
memmove(processedBuffer, processedBuffer+960*2, remainingDataLen);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if(levelMeter)
|
|
|
|
levelMeter->Update(reinterpret_cast<int16_t *>(data), len/2);
|
|
|
|
return len;
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2017-04-28 13:17:56 +02:00
|
|
|
void tgvoip::OpusDecoder::Start(){
|
2018-05-15 20:23:46 +02:00
|
|
|
if(!async)
|
|
|
|
return;
|
2017-02-02 17:24:40 +01:00
|
|
|
running=true;
|
2018-11-09 16:44:01 +01:00
|
|
|
thread=new Thread(std::bind(&tgvoip::OpusDecoder::RunThread, this));
|
2018-05-15 20:23:46 +02:00
|
|
|
thread->SetName("opus_decoder");
|
|
|
|
thread->SetMaxPriority();
|
|
|
|
thread->Start();
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
|
2017-04-28 13:17:56 +02:00
|
|
|
void tgvoip::OpusDecoder::Stop(){
|
2018-05-15 20:23:46 +02:00
|
|
|
if(!running || !async)
|
2017-02-02 17:24:40 +01:00
|
|
|
return;
|
|
|
|
running=false;
|
2018-05-15 20:23:46 +02:00
|
|
|
semaphore->Release();
|
|
|
|
thread->Join();
|
|
|
|
delete thread;
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
|
2018-11-09 16:44:01 +01:00
|
|
|
void tgvoip::OpusDecoder::RunThread(){
|
2017-02-02 17:24:40 +01:00
|
|
|
int i;
|
|
|
|
LOGI("decoder: packets per frame %d", packetsPerFrame);
|
|
|
|
while(running){
|
2018-05-15 20:23:46 +02:00
|
|
|
int playbackDuration=DecodeNextFrame();
|
|
|
|
for(i=0;i<playbackDuration/20;i++){
|
|
|
|
semaphore->Acquire();
|
2017-04-28 14:42:48 +02:00
|
|
|
if(!running){
|
|
|
|
LOGI("==== decoder exiting ====");
|
|
|
|
return;
|
|
|
|
}
|
2017-02-02 17:24:40 +01:00
|
|
|
unsigned char *buf=bufferPool->Get();
|
|
|
|
if(buf){
|
2018-05-15 20:23:46 +02:00
|
|
|
if(remainingDataLen>0){
|
2018-12-30 00:24:55 +01:00
|
|
|
for(effects::AudioEffect*& effect:postProcEffects){
|
|
|
|
effect->Process(reinterpret_cast<int16_t*>(processedBuffer+(PACKET_SIZE*i)), 960);
|
2017-09-07 07:39:33 +02:00
|
|
|
}
|
2017-05-21 16:50:23 +02:00
|
|
|
memcpy(buf, processedBuffer+(PACKET_SIZE*i), PACKET_SIZE);
|
2017-02-02 17:24:40 +01:00
|
|
|
}else{
|
2018-05-15 20:23:46 +02:00
|
|
|
//LOGE("Error decoding, result=%d", size);
|
2017-02-02 17:24:40 +01:00
|
|
|
memset(buf, 0, PACKET_SIZE);
|
|
|
|
}
|
|
|
|
decodedQueue->Put(buf);
|
|
|
|
}else{
|
|
|
|
LOGW("decoder: no buffers left!");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-05-15 20:23:46 +02:00
|
|
|
int tgvoip::OpusDecoder::DecodeNextFrame(){
|
|
|
|
int playbackDuration=0;
|
2018-06-04 21:37:43 +02:00
|
|
|
bool isEC=false;
|
|
|
|
size_t len=jitterBuffer->HandleOutput(buffer, 8192, 0, true, playbackDuration, isEC);
|
2018-05-15 20:23:46 +02:00
|
|
|
bool fec=false;
|
|
|
|
if(!len){
|
|
|
|
fec=true;
|
2018-06-04 21:37:43 +02:00
|
|
|
len=jitterBuffer->HandleOutput(buffer, 8192, 0, false, playbackDuration, isEC);
|
|
|
|
//if(len)
|
|
|
|
// LOGV("Trying FEC...");
|
2018-05-15 20:23:46 +02:00
|
|
|
}
|
|
|
|
int size;
|
|
|
|
if(len){
|
2018-06-04 21:37:43 +02:00
|
|
|
size=opus_decode(isEC ? ecDec : dec, buffer, len, (opus_int16 *) decodeBuffer, packetsPerFrame*960, fec ? 1 : 0);
|
2018-05-15 20:23:46 +02:00
|
|
|
consecutiveLostPackets=0;
|
2018-06-04 21:37:43 +02:00
|
|
|
if(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(prevWasEC ? ecDec : dec, NULL, 0, (opus_int16*)nextBuffer, packetsPerFrame*960, 0);
|
|
|
|
if(size){
|
|
|
|
int16_t* plcSamples=reinterpret_cast<int16_t*>(nextBuffer);
|
|
|
|
int16_t* samples=reinterpret_cast<int16_t*>(decodeBuffer);
|
|
|
|
constexpr float coeffs[]={0.999802, 0.995062, 0.984031, 0.966778, 0.943413, 0.914084, 0.878975, 0.838309, 0.792344,
|
|
|
|
0.741368, 0.685706, 0.625708, 0.561754, 0.494249, 0.423619, 0.350311, 0.274788, 0.197527, 0.119018, 0.039757};
|
|
|
|
for(int i=0;i<20;i++){
|
|
|
|
samples[i]=(int16_t)round((plcSamples[i]*coeffs[i]+(float)samples[i]*(1.0-coeffs[i])));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
prevWasEC=isEC;
|
|
|
|
prevLastSample=decodeBuffer[size-1];
|
2018-05-15 20:23:46 +02:00
|
|
|
}else{ // do packet loss concealment
|
|
|
|
consecutiveLostPackets++;
|
|
|
|
if(consecutiveLostPackets>2 && enableDTX){
|
|
|
|
silentPacketCount+=packetsPerFrame;
|
|
|
|
size=packetsPerFrame*960;
|
|
|
|
}else{
|
2018-06-04 21:37:43 +02:00
|
|
|
size=opus_decode(prevWasEC ? ecDec : dec, NULL, 0, (opus_int16 *) decodeBuffer, packetsPerFrame*960, 0);
|
2018-05-15 20:23:46 +02:00
|
|
|
//LOGV("PLC");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if(size<0)
|
|
|
|
LOGW("decoder: opus_decode error %d", size);
|
|
|
|
remainingDataLen=size;
|
|
|
|
if(playbackDuration==80){
|
|
|
|
processedBuffer=buffer;
|
|
|
|
audio::Resampler::Rescale60To80((int16_t*) decodeBuffer, (int16_t*) processedBuffer);
|
|
|
|
}else if(playbackDuration==40){
|
|
|
|
processedBuffer=buffer;
|
|
|
|
audio::Resampler::Rescale60To40((int16_t*) decodeBuffer, (int16_t*) processedBuffer);
|
|
|
|
}else{
|
|
|
|
processedBuffer=decodeBuffer;
|
|
|
|
}
|
|
|
|
return playbackDuration;
|
|
|
|
}
|
|
|
|
|
2017-02-02 17:24:40 +01:00
|
|
|
|
2017-04-28 13:17:56 +02:00
|
|
|
void tgvoip::OpusDecoder::SetFrameDuration(uint32_t duration){
|
2017-02-02 17:24:40 +01:00
|
|
|
frameDuration=duration;
|
2018-05-15 20:23:46 +02:00
|
|
|
packetsPerFrame=frameDuration/20;
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2018-06-04 21:37:43 +02:00
|
|
|
void tgvoip::OpusDecoder::SetJitterBuffer(std::shared_ptr<JitterBuffer> jitterBuffer){
|
2018-05-15 20:23:46 +02:00
|
|
|
this->jitterBuffer=jitterBuffer;
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
|
|
|
|
2018-05-15 20:23:46 +02:00
|
|
|
void tgvoip::OpusDecoder::SetDTX(bool enable){
|
|
|
|
enableDTX=enable;
|
|
|
|
}
|
2017-02-02 17:24:40 +01:00
|
|
|
|
2018-05-15 20:23:46 +02:00
|
|
|
void tgvoip::OpusDecoder::SetLevelMeter(AudioLevelMeter *levelMeter){
|
|
|
|
this->levelMeter=levelMeter;
|
2017-02-02 17:24:40 +01:00
|
|
|
}
|
2017-09-07 07:39:33 +02:00
|
|
|
|
2018-12-30 00:24:55 +01:00
|
|
|
void tgvoip::OpusDecoder::AddAudioEffect(effects::AudioEffect *effect){
|
2017-09-07 07:39:33 +02:00
|
|
|
postProcEffects.push_back(effect);
|
|
|
|
}
|
|
|
|
|
2018-12-30 00:24:55 +01:00
|
|
|
void tgvoip::OpusDecoder::RemoveAudioEffect(effects::AudioEffect *effect){
|
|
|
|
std::vector<effects::AudioEffect*>::iterator i=std::find(postProcEffects.begin(), postProcEffects.end(), effect);
|
2017-09-07 07:39:33 +02:00
|
|
|
if(i!=postProcEffects.end())
|
|
|
|
postProcEffects.erase(i);
|
|
|
|
}
|