- work to improve the decoding process - many problems fixes + however there are problems yet
This commit is contained in:
@@ -11,7 +11,7 @@ using namespace Audio;
|
||||
DataWindow::DataWindow()
|
||||
{
|
||||
mFilled = 0;
|
||||
mData = NULL;
|
||||
mData = nullptr;
|
||||
mCapacity = 0;
|
||||
}
|
||||
|
||||
@@ -166,6 +166,25 @@ void DataWindow::zero(int length)
|
||||
memset(mData, 0, mFilled);
|
||||
}
|
||||
|
||||
size_t DataWindow::moveTo(DataWindow& dst, size_t size)
|
||||
{
|
||||
Lock l(mMutex);
|
||||
|
||||
size_t avail = std::min(size, (size_t)filled());
|
||||
if (avail != 0)
|
||||
{
|
||||
dst.add(mData, avail);
|
||||
erase(avail);
|
||||
}
|
||||
return avail;
|
||||
}
|
||||
|
||||
std::chrono::milliseconds DataWindow::getTimeLength(int samplerate, int channels) const
|
||||
{
|
||||
Lock l(mMutex);
|
||||
return std::chrono::milliseconds(mFilled / sizeof(short) / channels / (samplerate / 1000));
|
||||
}
|
||||
|
||||
void DataWindow::makeStereoFromMono(DataWindow& dst, DataWindow& src)
|
||||
{
|
||||
Lock lockDst(dst.mMutex), lockSrc(src.mMutex);
|
||||
|
||||
@@ -11,9 +11,9 @@
|
||||
|
||||
namespace Audio
|
||||
{
|
||||
class DataWindow
|
||||
{
|
||||
public:
|
||||
class DataWindow
|
||||
{
|
||||
public:
|
||||
DataWindow();
|
||||
~DataWindow();
|
||||
|
||||
@@ -34,14 +34,17 @@ namespace Audio
|
||||
short shortAt(int index) const;
|
||||
void setShortAt(short value, int index);
|
||||
void zero(int length);
|
||||
size_t moveTo(DataWindow& dst, size_t size);
|
||||
|
||||
std::chrono::milliseconds getTimeLength(int samplerate, int channels) const;
|
||||
|
||||
static void makeStereoFromMono(DataWindow& dst, DataWindow& src);
|
||||
|
||||
protected:
|
||||
protected:
|
||||
mutable Mutex mMutex;
|
||||
char* mData;
|
||||
int mFilled;
|
||||
int mCapacity;
|
||||
};
|
||||
};
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -51,6 +51,11 @@ struct Format
|
||||
return float((milliseconds * mRate) / 500.0 * mChannels);
|
||||
}
|
||||
|
||||
size_t sizeFromTime(std::chrono::milliseconds ms) const
|
||||
{
|
||||
return sizeFromTime(ms.count());
|
||||
}
|
||||
|
||||
std::string toString()
|
||||
{
|
||||
char buffer[64];
|
||||
|
||||
@@ -50,13 +50,11 @@
|
||||
#define MT_MAXRTPPACKET 1500
|
||||
#define MT_DTMF_END_PACKETS 3
|
||||
|
||||
#define RTP_BUFFER_HIGH 0
|
||||
#define RTP_BUFFER_LOW 0
|
||||
#define RTP_BUFFER_PREBUFFER 0
|
||||
// Milliseconds before
|
||||
#define RTP_BUFFER_HIGH (2000)
|
||||
#define RTP_BUFFER_LOW (0)
|
||||
#define RTP_BUFFER_PREBUFFER (100)
|
||||
|
||||
// #define RTP_BUFFER_HIGH 160
|
||||
// #define RTP_BUFFER_LOW 10
|
||||
// #define RTP_BUFFER_PREBUFFER 160
|
||||
#define RTP_DECODED_CAPACITY 2048
|
||||
|
||||
#define DEFAULT_SUBSCRIPTION_TIME 1200
|
||||
|
||||
@@ -31,30 +31,30 @@ const uint16_t amrwb_framelenbits[10] =
|
||||
|
||||
struct AmrPayloadInfo
|
||||
{
|
||||
const uint8_t* mPayload;
|
||||
int mPayloadLength;
|
||||
bool mOctetAligned;
|
||||
bool mInterleaving;
|
||||
bool mWideband;
|
||||
uint64_t mCurrentTimestamp;
|
||||
const uint8_t* mPayload = nullptr;
|
||||
int mPayloadLength = 0;
|
||||
bool mOctetAligned = false;
|
||||
bool mInterleaving = false;
|
||||
bool mWideband = false;
|
||||
uint64_t mCurrentTimestamp = 0;
|
||||
};
|
||||
|
||||
|
||||
struct AmrFrame
|
||||
{
|
||||
uint8_t mFrameType;
|
||||
uint8_t mMode;
|
||||
bool mGoodQuality;
|
||||
uint64_t mTimestamp;
|
||||
uint8_t mFrameType = 0;
|
||||
uint8_t mMode = 0;
|
||||
bool mGoodQuality = false;
|
||||
uint64_t mTimestamp = 0;
|
||||
std::shared_ptr<ByteBuffer> mData;
|
||||
uint8_t mSTI;
|
||||
uint8_t mSTI = 0;
|
||||
};
|
||||
|
||||
struct AmrPayload
|
||||
{
|
||||
uint8_t mCodeModeRequest;
|
||||
uint8_t mCodeModeRequest = 0;
|
||||
std::vector<AmrFrame> mFrames;
|
||||
bool mDiscardPacket;
|
||||
bool mDiscardPacket = false;
|
||||
};
|
||||
|
||||
// ARM RTP payload has next structure
|
||||
@@ -148,10 +148,10 @@ static AmrPayload parseAmrPayload(AmrPayloadInfo& input, size_t& cngCounter)
|
||||
continue;
|
||||
}
|
||||
|
||||
if (input.mWideband && f.mMode == 0xFF /* CNG */)
|
||||
{
|
||||
int a = 1;
|
||||
}
|
||||
// if (input.mWideband && f.mMode == 0xFF /* CNG */)
|
||||
// {
|
||||
// int a = 1;
|
||||
// }
|
||||
|
||||
if (input.mWideband && f.mFrameType == 15)
|
||||
{
|
||||
@@ -628,21 +628,30 @@ int AmrWbCodec::decodePlain(std::span<const uint8_t> input, std::span<uint8_t> o
|
||||
return 0;
|
||||
}
|
||||
|
||||
// Check for output buffer capacity
|
||||
if (output.size() < (int)ap.mFrames.size() * pcmLength())
|
||||
// Find the required output capacity
|
||||
size_t capacity = 0;
|
||||
for (AmrFrame& frame: ap.mFrames)
|
||||
capacity += frame.mMode == 0xFF /* CNG */ ? pcmLength() * 8 : pcmLength();
|
||||
|
||||
if (output.size() < capacity)
|
||||
return 0;
|
||||
|
||||
short* dataOut = (short*)output.data();
|
||||
size_t dataOutSizeInBytes = 0;
|
||||
for (AmrFrame& frame: ap.mFrames)
|
||||
{
|
||||
memset(dataOut, 0, static_cast<size_t>(pcmLength()));
|
||||
size_t frameOutputSize = frame.mMode == 0xFF ? pcmLength() * 8 : pcmLength();
|
||||
memset(dataOut, 0, frameOutputSize);
|
||||
|
||||
if (frame.mData)
|
||||
{
|
||||
if (frame.mMode == 0xFF)
|
||||
{
|
||||
// int bp = 1;
|
||||
}
|
||||
D_IF_decode(mDecoderCtx, (const unsigned char*)frame.mData->data(), (short*)dataOut, 0);
|
||||
dataOut += pcmLength() / 2;
|
||||
dataOutSizeInBytes += pcmLength();
|
||||
dataOut += frameOutputSize / 2;
|
||||
dataOutSizeInBytes += frameOutputSize;
|
||||
}
|
||||
}
|
||||
return dataOutSizeInBytes;
|
||||
|
||||
@@ -635,10 +635,10 @@ int IlbcCodec::samplerate()
|
||||
return 8000;
|
||||
}
|
||||
|
||||
int IlbcCodec::encode(const void *input, int inputBytes, void* outputBuffer, int outputCapacity)
|
||||
Codec::EncodeResult IlbcCodec::encode(const void *input, int inputBytes, void* outputBuffer, int outputCapacity)
|
||||
{
|
||||
if (inputBytes % pcmLength())
|
||||
return 0;
|
||||
return {};
|
||||
|
||||
// Declare the data input pointer
|
||||
short *dataIn = (short *)input;
|
||||
@@ -657,10 +657,10 @@ int IlbcCodec::encode(const void *input, int inputBytes, void* outputBuffer, int
|
||||
dataOut += rtpLength();
|
||||
}
|
||||
|
||||
return frames * rtpLength();
|
||||
return {frames * rtpLength()};
|
||||
}
|
||||
|
||||
int IlbcCodec::decode(const void* input, int inputBytes, void* output, int outputCapacity)
|
||||
Codec::DecodeResult IlbcCodec::decode(const void* input, int inputBytes, void* output, int outputCapacity)
|
||||
{
|
||||
unsigned frames = inputBytes / rtpLength();
|
||||
|
||||
@@ -675,12 +675,12 @@ int IlbcCodec::decode(const void* input, int inputBytes, void* output, int outpu
|
||||
dataOut += pcmLength() / 2;
|
||||
}
|
||||
|
||||
return frames * pcmLength();
|
||||
return {frames * pcmLength()};
|
||||
}
|
||||
|
||||
int IlbcCodec::plc(int lostFrames, void* output, int outputCapacity)
|
||||
int IlbcCodec::plc(int lostFrames, std::span<uint8_t> output)
|
||||
{
|
||||
return 2 * WebRtcIlbcfix_DecodePlc(mDecoderCtx, (WebRtc_Word16*)output, lostFrames);
|
||||
return sizeof(short) * WebRtcIlbcfix_DecodePlc(mDecoderCtx, (WebRtc_Word16*)output.data(), lostFrames);
|
||||
}
|
||||
|
||||
// --- IlbcFactory ---
|
||||
|
||||
@@ -58,9 +58,10 @@ public:
|
||||
int frameTime() override;
|
||||
int samplerate() override;
|
||||
int channels() override;
|
||||
int encode(const void* input, int inputBytes, void* output, int outputCapacity) override;
|
||||
int decode(const void* input, int inputBytes, void* output, int outputCapacity) override;
|
||||
int plc(int lostFrames, void* output, int outputCapacity) override;
|
||||
|
||||
EncodeResult encode(std::span<const uint8_t> input, std::span<uint8_t> output) override;
|
||||
DecodeResult decode(std::span<const uint8_t> input, std::span<uint8_t> output) override;
|
||||
size_t plc(int lostFrames, std::span<uint8_t> output) override;
|
||||
};
|
||||
|
||||
class OpusCodec: public Codec
|
||||
@@ -112,9 +113,10 @@ public:
|
||||
int frameTime();
|
||||
int samplerate();
|
||||
int channels();
|
||||
int encode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int decode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int plc(int lostFrames, void* output, int outputCapacity);
|
||||
|
||||
EncodeResult encode(std::span<const uint8_t> input, std::span<uint8_t> output);
|
||||
DecodeResult decode(std::span<const uint8_t> input, std::span<uint8_t> output);
|
||||
size_t plc(int lostFrames, std::span<uint8_t> output);
|
||||
};
|
||||
|
||||
|
||||
@@ -146,14 +148,15 @@ public:
|
||||
|
||||
IlbcCodec(int packetTime);
|
||||
virtual ~IlbcCodec();
|
||||
const char* name();
|
||||
int pcmLength();
|
||||
int rtpLength();
|
||||
int frameTime();
|
||||
int samplerate();
|
||||
int encode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int decode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int plc(int lostFrames, void* output, int outputCapacity);
|
||||
const char* name() override;
|
||||
int pcmLength() override;
|
||||
int rtpLength() override;
|
||||
int frameTime() override;
|
||||
int samplerate() override;
|
||||
|
||||
EncodeResult encode(std::span<const uint8_t> input, std::span<uint8_t> output) override;
|
||||
DecodeResult decode(std::span<const uint8_t> input, std::span<uint8_t> output) override;
|
||||
size_t plc(int lostFrames, std::span<uint8_t> output) override;
|
||||
};
|
||||
|
||||
class G711Codec: public Codec
|
||||
@@ -186,15 +189,15 @@ public:
|
||||
G711Codec(int type);
|
||||
~G711Codec();
|
||||
|
||||
const char* name();
|
||||
int pcmLength();
|
||||
int frameTime();
|
||||
int rtpLength();
|
||||
int samplerate();
|
||||
const char* name() override;
|
||||
int pcmLength() override;
|
||||
int frameTime() override;
|
||||
int rtpLength() override;
|
||||
int samplerate() override;
|
||||
|
||||
int encode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int decode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int plc(int lostSamples, void* output, int outputCapacity);
|
||||
EncodeResult encode(std::span<const uint8_t> input, std::span<uint8_t> output) override;
|
||||
DecodeResult decode(std::span<const uint8_t> input, std::span<uint8_t> output) override;
|
||||
size_t plc(int lostSamples, std::span<uint8_t> output) override ;
|
||||
|
||||
protected:
|
||||
int mType; /// Determines if it is u-law or a-law codec. Its value is ALaw or ULaw.
|
||||
@@ -237,15 +240,15 @@ public:
|
||||
IsacCodec(int sampleRate);
|
||||
~IsacCodec();
|
||||
|
||||
const char* name();
|
||||
int pcmLength();
|
||||
int rtpLength();
|
||||
int frameTime();
|
||||
int samplerate();
|
||||
const char* name() override;
|
||||
int pcmLength() override;
|
||||
int rtpLength() override;
|
||||
int frameTime() override;
|
||||
int samplerate() override;
|
||||
|
||||
int encode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int decode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int plc(int lostFrames, void* output, int outputCapacity);
|
||||
EncodeResult encode(std::span<const uint8_t> input, std::span<uint8_t> output) override;
|
||||
DecodeResult decode(std::span<const uint8_t> input, std::span<uint8_t> output) override;
|
||||
size_t plc(int lostFrames, std::span<uint8_t> output) override;
|
||||
};
|
||||
|
||||
|
||||
@@ -311,11 +314,11 @@ public:
|
||||
/*! Destructor. */
|
||||
virtual ~GsmCodec();
|
||||
|
||||
const char* name();
|
||||
int pcmLength();
|
||||
int rtpLength();
|
||||
int frameTime();
|
||||
int samplerate();
|
||||
const char* name() override;
|
||||
int pcmLength() override;
|
||||
int rtpLength() override;
|
||||
int frameTime() override;
|
||||
int samplerate() override;
|
||||
|
||||
int encode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
int decode(const void* input, int inputBytes, void* output, int outputCapacity);
|
||||
|
||||
@@ -115,7 +115,7 @@ bool SequenceSort(const std::shared_ptr<RtpBuffer::Packet>& p1, const std::share
|
||||
return p1->rtp()->GetExtendedSequenceNumber() < p2->rtp()->GetExtendedSequenceNumber();
|
||||
}
|
||||
|
||||
std::shared_ptr<RtpBuffer::Packet> RtpBuffer::add(std::shared_ptr<jrtplib::RTPPacket> packet, std::chrono::milliseconds timelength, int rate)
|
||||
std::shared_ptr<RtpBuffer::Packet> RtpBuffer::add(const std::shared_ptr<jrtplib::RTPPacket>& packet, std::chrono::milliseconds timelength, int rate)
|
||||
{
|
||||
if (!packet)
|
||||
return std::shared_ptr<Packet>();
|
||||
@@ -191,12 +191,11 @@ std::shared_ptr<RtpBuffer::Packet> RtpBuffer::add(std::shared_ptr<jrtplib::RTPPa
|
||||
return std::shared_ptr<Packet>();
|
||||
}
|
||||
|
||||
RtpBuffer::FetchResult RtpBuffer::fetch(ResultList& rl)
|
||||
RtpBuffer::FetchResult RtpBuffer::fetch()
|
||||
{
|
||||
Lock l(mGuard);
|
||||
|
||||
FetchResult result = FetchResult::NoPacket;
|
||||
rl.clear();
|
||||
FetchResult result;
|
||||
|
||||
// See if there is enough information in buffer
|
||||
auto total = findTimelength();
|
||||
@@ -217,10 +216,10 @@ RtpBuffer::FetchResult RtpBuffer::fetch(ResultList& rl)
|
||||
mStat.mPacketDropped++;
|
||||
}
|
||||
|
||||
if (total < mLow)
|
||||
if (total < mLow || total == 0ms)
|
||||
{
|
||||
// Still not prebuffered
|
||||
result = FetchResult::NoPacket;
|
||||
result = {FetchResult::Status::NoPacket};
|
||||
}
|
||||
else
|
||||
{
|
||||
@@ -228,8 +227,8 @@ RtpBuffer::FetchResult RtpBuffer::fetch(ResultList& rl)
|
||||
{
|
||||
if (mPacketList.empty())
|
||||
{
|
||||
result = FetchResult::NoPacket;
|
||||
// Don't increase counter of lost packets here; maybe it is DTX
|
||||
result = {FetchResult::Status::NoPacket};
|
||||
}
|
||||
else
|
||||
{
|
||||
@@ -237,7 +236,6 @@ RtpBuffer::FetchResult RtpBuffer::fetch(ResultList& rl)
|
||||
auto& packet = *mPacketList.front();
|
||||
uint32_t seqno = packet.rtp()->GetExtendedSequenceNumber();
|
||||
|
||||
|
||||
// Gap between new packet and previous on
|
||||
int gap = (int64_t)seqno - (int64_t)*mLastSeqno - 1;
|
||||
gap = std::min(gap, 127);
|
||||
@@ -255,16 +253,15 @@ RtpBuffer::FetchResult RtpBuffer::fetch(ResultList& rl)
|
||||
|
||||
mLastSeqno = *mLastSeqno + 1; // As we deal with the audio gap - return the silence and increase last seqno
|
||||
|
||||
result = FetchResult::Gap;
|
||||
result = {FetchResult::Status::Gap};
|
||||
}
|
||||
else
|
||||
{
|
||||
result = FetchResult::RegularPacket;
|
||||
rl.push_back(mPacketList.front());
|
||||
result = {FetchResult::Status::RegularPacket, mPacketList.front()};
|
||||
|
||||
// Save last returned normal packet
|
||||
mFetchedPacket = mPacketList.front();
|
||||
mLastSeqno = mPacketList.front()->rtp()->GetExtendedSequenceNumber();
|
||||
mFetchedPacket = result.mPacket;
|
||||
mLastSeqno = result.mPacket->rtp()->GetExtendedSequenceNumber();
|
||||
|
||||
// Remove returned packet from the list
|
||||
mPacketList.erase(mPacketList.begin());
|
||||
@@ -277,14 +274,11 @@ RtpBuffer::FetchResult RtpBuffer::fetch(ResultList& rl)
|
||||
if (findTimelength() >= mPrebuffer && !mPacketList.empty())
|
||||
{
|
||||
// Normal packet will be returned
|
||||
result = FetchResult::RegularPacket;
|
||||
|
||||
// Put it to output list
|
||||
rl.push_back(mPacketList.front());
|
||||
result = {FetchResult::Status::RegularPacket, mPacketList.front()};
|
||||
|
||||
// Remember returned packet
|
||||
mFetchedPacket = mPacketList.front();
|
||||
mLastSeqno = mPacketList.front()->rtp()->GetExtendedSequenceNumber();
|
||||
mFetchedPacket = result.mPacket;
|
||||
mLastSeqno = result.mPacket->rtp()->GetExtendedSequenceNumber();
|
||||
|
||||
// Remove returned packet from buffer list
|
||||
mPacketList.erase(mPacketList.begin());
|
||||
@@ -292,12 +286,12 @@ RtpBuffer::FetchResult RtpBuffer::fetch(ResultList& rl)
|
||||
else
|
||||
{
|
||||
ICELogMedia(<< "Jitter buffer was not prebuffered yet; resulting no packet");
|
||||
result = FetchResult::NoPacket;
|
||||
result = {FetchResult::Status::NoPacket};
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if (result != FetchResult::NoPacket)
|
||||
if (result.mStatus != FetchResult::Status::NoPacket)
|
||||
mReturnedCounter++;
|
||||
|
||||
return result;
|
||||
@@ -333,8 +327,7 @@ Receiver::~Receiver()
|
||||
|
||||
//-------------- AudioReceiver ----------------
|
||||
AudioReceiver::AudioReceiver(const CodecList::Settings& settings, MT::Statistics &stat)
|
||||
:Receiver(stat), mBuffer(stat), mCodecSettings(settings),
|
||||
mCodecList(settings)
|
||||
:Receiver(stat), mBuffer(stat), mCodecSettings(settings), mCodecList(settings)
|
||||
{
|
||||
// Init resamplers
|
||||
mResampler8.start(AUDIO_CHANNELS, 8000, AUDIO_SAMPLERATE);
|
||||
@@ -346,6 +339,8 @@ AudioReceiver::AudioReceiver(const CodecList::Settings& settings, MT::Statistics
|
||||
mCodecList.setSettings(settings);
|
||||
mCodecList.fillCodecMap(mCodecMap);
|
||||
|
||||
mAvailable.setCapacity(AUDIO_SAMPLERATE * sizeof(short));
|
||||
|
||||
#if defined(DUMP_DECODED)
|
||||
mDecodedDump = std::make_shared<Audio::WavFileWriter>();
|
||||
mDecodedDump->open("decoded.wav", 8000 /*G711*/, AUDIO_CHANNELS);
|
||||
@@ -559,11 +554,15 @@ AudioReceiver::DecodeResult AudioReceiver::decodeGapTo(Audio::DataWindow& output
|
||||
|
||||
mDecodedLength = mResampledLength = 0;
|
||||
if (mCngPacket && mCodec)
|
||||
{
|
||||
if (mCngPacket->rtp()->GetPayloadType() == 13)
|
||||
{
|
||||
// Synthesize comfort noise. It will be done on AUDIO_SAMPLERATE rate directly to mResampledFrame buffer.
|
||||
// Do not forget to send this noise to analysis
|
||||
mDecodedLength = mCngDecoder.produce(mCodec->samplerate(), mLastPacketTimeLength,
|
||||
reinterpret_cast<short*>(mDecodedFrame), false);
|
||||
mDecodedLength = mCngDecoder.produce(mCodec->samplerate(), mLastPacketTimeLength, reinterpret_cast<short*>(mDecodedFrame), false);
|
||||
}
|
||||
else
|
||||
decodePacketTo(output, options, mCngPacket);
|
||||
}
|
||||
else
|
||||
if (mCodec && mFrameCount && !mCodecSettings.mSkipDecode)
|
||||
@@ -594,19 +593,19 @@ AudioReceiver::DecodeResult AudioReceiver::decodeGapTo(Audio::DataWindow& output
|
||||
return {.mStatus = DecodeResult::Status::Skip};
|
||||
}
|
||||
|
||||
AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& output, DecodeOptions options, const RtpBuffer::ResultList& rl)
|
||||
AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& output, DecodeOptions options, const std::shared_ptr<RtpBuffer::Packet>& packet)
|
||||
{
|
||||
if (!packet || !packet->rtp())
|
||||
return {DecodeResult::Status::Skip};
|
||||
|
||||
DecodeResult result = {.mStatus = DecodeResult::Status::Skip};
|
||||
auto& rtp = *packet->rtp(); // Syntax sugar
|
||||
|
||||
mFailedCount = 0;
|
||||
for (const std::shared_ptr<RtpBuffer::Packet>& p: rl)
|
||||
{
|
||||
assert(p);
|
||||
|
||||
// Check if we need to emit silence or CNG - previously CNG packet was detected. Emit CNG audio here if needed.
|
||||
if (mLastPacketTimestamp && mLastPacketTimeLength && mCodec)
|
||||
{
|
||||
int units = p->rtp()->GetTimestamp() - *mLastPacketTimestamp;
|
||||
int units = rtp.GetTimestamp() - *mLastPacketTimestamp;
|
||||
int milliseconds = units / (mCodec->samplerate() / 1000);
|
||||
if (milliseconds > mLastPacketTimeLength)
|
||||
{
|
||||
@@ -619,17 +618,16 @@ AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& out
|
||||
}
|
||||
}
|
||||
|
||||
mLastPacketTimestamp = p->rtp()->GetTimestamp();
|
||||
mLastPacketTimestamp = rtp.GetTimestamp();
|
||||
|
||||
// Find codec by payload type
|
||||
int ptype = p->rtp()->GetPayloadType();
|
||||
int ptype = rtp.GetPayloadType();
|
||||
|
||||
// Look into mCodecMap if exists
|
||||
auto codecIter = mCodecMap.find(ptype);
|
||||
if (codecIter == mCodecMap.end())
|
||||
return {};
|
||||
|
||||
|
||||
if (!codecIter->second)
|
||||
codecIter->second = mCodecList.createCodecByPayloadType(ptype);
|
||||
|
||||
@@ -640,18 +638,17 @@ AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& out
|
||||
result.mSamplerate = mCodec->samplerate();
|
||||
|
||||
// Check if it is CNG packet
|
||||
if ((ptype == 0 || ptype == 8) && p->rtp()->GetPayloadLength() >= 1 && p->rtp()->GetPayloadLength() <= 6)
|
||||
if (((ptype == 0 || ptype == 8) && rtp.GetPayloadLength() >= 1 && rtp.GetPayloadLength() <= 6) || rtp.GetPayloadType() == 13)
|
||||
{
|
||||
if (options.mSkipDecode)
|
||||
mDecodedLength = 0;
|
||||
else
|
||||
{
|
||||
mCngPacket = p->rtp();
|
||||
mCngDecoder.decode3389(p->rtp()->GetPayloadData(), p->rtp()->GetPayloadLength());
|
||||
mCngPacket = packet;
|
||||
mCngDecoder.decode3389(rtp.GetPayloadData(), rtp.GetPayloadLength());
|
||||
|
||||
// Emit CNG mLastPacketLength milliseconds
|
||||
mDecodedLength = mCngDecoder.produce(mCodec->samplerate(), mLastPacketTimeLength,
|
||||
(short*)mDecodedFrame, true);
|
||||
mDecodedLength = mCngDecoder.produce(mCodec->samplerate(), mLastPacketTimeLength, (short*)mDecodedFrame, true);
|
||||
if (mDecodedLength)
|
||||
processDecoded(output, options);
|
||||
}
|
||||
@@ -664,7 +661,7 @@ AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& out
|
||||
|
||||
// Handle here regular RTP packets
|
||||
// Check if payload length is ok
|
||||
size_t payload_length = p->rtp()->GetPayloadLength();
|
||||
size_t payload_length = rtp.GetPayloadLength();
|
||||
size_t rtp_frame_length = mCodec->rtpLength();
|
||||
|
||||
int tail = rtp_frame_length ? payload_length % rtp_frame_length : 0;
|
||||
@@ -672,8 +669,8 @@ AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& out
|
||||
if (!tail)
|
||||
{
|
||||
// Find number of frames
|
||||
mFrameCount = mCodec->rtpLength() ? p->rtp()->GetPayloadLength() / mCodec->rtpLength() : 1;
|
||||
int frameLength = mCodec->rtpLength() ? mCodec->rtpLength() : (int)p->rtp()->GetPayloadLength();
|
||||
mFrameCount = mCodec->rtpLength() ? rtp.GetPayloadLength() / mCodec->rtpLength() : 1;
|
||||
int frameLength = mCodec->rtpLength() ? mCodec->rtpLength() : (int)rtp.GetPayloadLength();
|
||||
|
||||
// Save last packet time length
|
||||
mLastPacketTimeLength = mFrameCount * mCodec->frameTime();
|
||||
@@ -686,8 +683,7 @@ AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& out
|
||||
else
|
||||
{
|
||||
// Decode frame by frame
|
||||
mDecodedLength = mCodec->decode(p->rtp()->GetPayloadData() + i * mCodec->rtpLength(),
|
||||
frameLength, mDecodedFrame, sizeof mDecodedFrame);
|
||||
mDecodedLength = mCodec->decode(rtp.GetPayloadData() + i * mCodec->rtpLength(), frameLength, mDecodedFrame, sizeof mDecodedFrame);
|
||||
if (mDecodedLength > 0)
|
||||
processDecoded(output, options);
|
||||
}
|
||||
@@ -704,13 +700,21 @@ AudioReceiver::DecodeResult AudioReceiver::decodePacketTo(Audio::DataWindow& out
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
return result;
|
||||
}
|
||||
|
||||
AudioReceiver::DecodeResult AudioReceiver::decodeEmptyTo(Audio::DataWindow& output, DecodeOptions options)
|
||||
{
|
||||
// No packet available in jitter buffer - just increase the counter for now
|
||||
// No packet available at all (and no previous CNG packet) - so return the silence
|
||||
if (options.mElapsed != 0ms && mCodec)
|
||||
{
|
||||
Audio::Format fmt = options.mResampleToMainRate ? Audio::Format(AUDIO_SAMPLERATE, 1) : mCodec->getAudioFormat();
|
||||
// Emit silence if codec information is available - it is to properly handle the gaps
|
||||
auto avail = output.getTimeLength(fmt.rate(), fmt.channels());
|
||||
if (options.mElapsed > avail)
|
||||
mAvailable.addZero(fmt.sizeFromTime(options.mElapsed - avail));
|
||||
}
|
||||
|
||||
mFailedCount++;
|
||||
return {.mStatus = DecodeResult::Status::Skip};
|
||||
}
|
||||
@@ -719,32 +723,71 @@ AudioReceiver::DecodeResult AudioReceiver::getAudioTo(Audio::DataWindow& output,
|
||||
{
|
||||
DecodeResult result = {.mStatus = DecodeResult::Status::Skip};
|
||||
|
||||
size_t initialOffset = output.filled(); // Size in bytes
|
||||
auto produced = 0ms;
|
||||
if (mAvailable.filled() && mCodec && options.mElapsed != 0ms)
|
||||
{
|
||||
Audio::Format fmt = options.mResampleToMainRate ? Audio::Format(AUDIO_SAMPLERATE, 1) : mCodec->getAudioFormat();
|
||||
auto initiallyAvailable = mCodec ? mAvailable.getTimeLength(fmt.rate(), fmt.channels()) : 0ms;
|
||||
if (initiallyAvailable != 0ms)
|
||||
{
|
||||
std::chrono::milliseconds resultTime = std::min(initiallyAvailable, options.mElapsed);
|
||||
auto resultLen = fmt.sizeFromTime(resultTime);
|
||||
mAvailable.moveTo(output, resultLen);
|
||||
produced += resultTime;
|
||||
|
||||
// Maybe request is satisfied ?
|
||||
if (produced >= options.mElapsed)
|
||||
return {.mStatus = DecodeResult::Status::Ok, .mSamplerate = fmt.rate(), .mChannels = fmt.channels()};
|
||||
}
|
||||
}
|
||||
|
||||
std::chrono::milliseconds decoded = 0ms;
|
||||
do
|
||||
{
|
||||
// Get next packet from buffer
|
||||
RtpBuffer::ResultList rl;
|
||||
RtpBuffer::FetchResult fr = mBuffer.fetch(rl);
|
||||
switch (fr)
|
||||
RtpBuffer::FetchResult fr = mBuffer.fetch();
|
||||
// ICELogDebug(<< fr.toString() << " " << mBuffer.findTimelength());
|
||||
|
||||
switch (fr.mStatus)
|
||||
{
|
||||
case RtpBuffer::FetchResult::Gap: result = decodeGapTo(output, options); break;
|
||||
case RtpBuffer::FetchResult::NoPacket: result = decodeEmptyTo(output, options); break;
|
||||
case RtpBuffer::FetchResult::RegularPacket: result = decodePacketTo(output, options, rl); break;
|
||||
case RtpBuffer::FetchResult::Status::Gap: result = decodeGapTo(mAvailable, options); break;
|
||||
case RtpBuffer::FetchResult::Status::NoPacket: result = decodeEmptyTo(mAvailable, options); break;
|
||||
case RtpBuffer::FetchResult::Status::RegularPacket: result = decodePacketTo(mAvailable, options, fr.mPacket); break;
|
||||
default:
|
||||
assert(0);
|
||||
}
|
||||
|
||||
size_t available = output.filled() - initialOffset;
|
||||
if (!available)
|
||||
break;
|
||||
initialOffset = output.filled();
|
||||
// Was there decoding at all ?
|
||||
if (!mCodec)
|
||||
break; // No sense to continue - we have no information at all
|
||||
|
||||
// ToDo: calculate how much milliseconds was decoded
|
||||
int samplerate = options.mResampleToMainRate ? AUDIO_SAMPLERATE : result.mSamplerate;
|
||||
decoded += std::chrono::milliseconds(available / sizeof(short) / (samplerate / 1000));
|
||||
Audio::Format fmt = options.mResampleToMainRate ? Audio::Format(AUDIO_SAMPLERATE, 1) : mCodec->getAudioFormat();
|
||||
result.mSamplerate = fmt.rate();
|
||||
result.mChannels = fmt.channels();
|
||||
|
||||
// Have we anything interesting in the buffer ?
|
||||
auto bufferAvailable = mAvailable.getTimeLength(fmt.rate(), fmt.channels());
|
||||
if (bufferAvailable == 0ms)
|
||||
break; // No sense to continue - decoding / CNG / PLC stopped totally
|
||||
|
||||
// How much data should be moved to result buffer ?
|
||||
if (options.mElapsed != 0ms)
|
||||
{
|
||||
std::chrono::milliseconds resultTime = std::min(bufferAvailable, options.mElapsed - produced);
|
||||
auto resultLen = fmt.sizeFromTime(resultTime);
|
||||
mAvailable.moveTo(output, resultLen);
|
||||
produced += resultTime;
|
||||
}
|
||||
while (decoded < options.mElapsed);
|
||||
else
|
||||
mAvailable.moveTo(output, mAvailable.filled());
|
||||
|
||||
decoded += bufferAvailable;
|
||||
}
|
||||
while (produced < options.mElapsed);
|
||||
|
||||
if (produced != 0ms)
|
||||
result.mStatus = DecodeResult::Status::Ok;
|
||||
|
||||
// Time statistics
|
||||
if (result.mStatus == DecodeResult::Status::Ok)
|
||||
|
||||
@@ -28,13 +28,6 @@ using jrtplib::RTPPacket;
|
||||
class RtpBuffer
|
||||
{
|
||||
public:
|
||||
enum class FetchResult
|
||||
{
|
||||
RegularPacket,
|
||||
Gap,
|
||||
NoPacket
|
||||
};
|
||||
|
||||
// Owns rtp packet data
|
||||
class Packet
|
||||
{
|
||||
@@ -59,6 +52,29 @@ public:
|
||||
std::chrono::microseconds mTimestamp = 0us;
|
||||
};
|
||||
|
||||
struct FetchResult
|
||||
{
|
||||
enum class Status
|
||||
{
|
||||
RegularPacket,
|
||||
Gap,
|
||||
NoPacket
|
||||
};
|
||||
|
||||
Status mStatus = Status::NoPacket;
|
||||
std::shared_ptr<Packet> mPacket;
|
||||
|
||||
std::string toString() const
|
||||
{
|
||||
switch (mStatus)
|
||||
{
|
||||
case Status::RegularPacket: return "packet";
|
||||
case Status::Gap: return "gap";
|
||||
case Status::NoPacket: return "empty";
|
||||
}
|
||||
}
|
||||
};
|
||||
|
||||
RtpBuffer(Statistics& stat);
|
||||
~RtpBuffer();
|
||||
|
||||
@@ -81,12 +97,12 @@ public:
|
||||
int getCount() const;
|
||||
|
||||
// Returns false if packet was not add - maybe too old or too new or duplicate
|
||||
std::shared_ptr<Packet> add(std::shared_ptr<RTPPacket> packet, std::chrono::milliseconds timelength, int rate);
|
||||
std::shared_ptr<Packet> add(const std::shared_ptr<RTPPacket>& packet, std::chrono::milliseconds timelength, int rate);
|
||||
|
||||
typedef std::vector<std::shared_ptr<Packet>> ResultList;
|
||||
typedef std::shared_ptr<ResultList> PResultList;
|
||||
|
||||
FetchResult fetch(ResultList& rl);
|
||||
FetchResult fetch();
|
||||
|
||||
protected:
|
||||
unsigned mSsrc = 0;
|
||||
@@ -133,15 +149,6 @@ public:
|
||||
// Lifetime of pointer to codec is limited by lifetime of AudioReceiver (it is container).
|
||||
bool add(const std::shared_ptr<jrtplib::RTPPacket>& p, Codec** codec = nullptr);
|
||||
|
||||
// Returns false when there is no rtp data from jitter
|
||||
/*enum DecodeOptions
|
||||
{
|
||||
DecodeOptions_ResampleToMainRate = 0,
|
||||
DecodeOptions_DontResample = 1,
|
||||
DecodeOptions_FillCngGap = 2,
|
||||
DecodeOptions_SkipDecode = 4
|
||||
};*/
|
||||
|
||||
struct DecodeOptions
|
||||
{
|
||||
bool mResampleToMainRate = true; // Resample all decoded audio to AUDIO_SAMPLERATE
|
||||
@@ -187,11 +194,14 @@ protected:
|
||||
CodecList::Settings mCodecSettings;
|
||||
CodecList mCodecList;
|
||||
JitterStatistics mJitterStats;
|
||||
std::shared_ptr<jrtplib::RTPPacket> mCngPacket;
|
||||
std::shared_ptr<RtpBuffer::Packet> mCngPacket;
|
||||
CngDecoder mCngDecoder;
|
||||
size_t mDTXSamplesToEmit = 0; // How much silence (or CNG) should be emited before next RTP packet gets into the action
|
||||
|
||||
// Buffer to hold decoded data
|
||||
// Already decoded data that can be retrieved without actual decoding - it may happen because of getAudioTo() may be limited by time interval
|
||||
Audio::DataWindow mAvailable;
|
||||
|
||||
// Temporary buffer to hold decoded data (it is better than allocate data on stack)
|
||||
int16_t mDecodedFrame[MT_MAX_DECODEBUFFER];
|
||||
size_t mDecodedLength = 0;
|
||||
|
||||
@@ -208,7 +218,10 @@ protected:
|
||||
std::optional<uint32_t> mLastPacketTimestamp;
|
||||
|
||||
int mFailedCount = 0;
|
||||
Audio::Resampler mResampler8, mResampler16, mResampler32, mResampler48;
|
||||
Audio::Resampler mResampler8,
|
||||
mResampler16,
|
||||
mResampler32,
|
||||
mResampler48;
|
||||
|
||||
Audio::PWavFileWriter mDecodedDump;
|
||||
|
||||
@@ -229,7 +242,7 @@ protected:
|
||||
void updateAmrCodecStats(Codec* c);
|
||||
|
||||
DecodeResult decodeGapTo(Audio::DataWindow& output, DecodeOptions options);
|
||||
DecodeResult decodePacketTo(Audio::DataWindow& output, DecodeOptions options, const RtpBuffer::ResultList& rl);
|
||||
DecodeResult decodePacketTo(Audio::DataWindow& output, DecodeOptions options, const std::shared_ptr<RtpBuffer::Packet>& p);
|
||||
DecodeResult decodeEmptyTo(Audio::DataWindow& output, DecodeOptions options);
|
||||
};
|
||||
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
/* Copyright(C) 2007-2014 VoIP objects (voipobjects.com)
|
||||
/* Copyright(C) 2007-2026 VoIP objects (voipobjects.com)
|
||||
* This Source Code Form is subject to the terms of the Mozilla Public
|
||||
* License, v. 2.0. If a copy of the MPL was not distributed with this
|
||||
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
|
||||
|
||||
@@ -10,7 +10,7 @@
|
||||
#include "../helper/HL_Types.h"
|
||||
#include <map>
|
||||
#include "../helper/HL_Pointer.h"
|
||||
|
||||
#include "../audio/Audio_Interface.h"
|
||||
|
||||
namespace MT
|
||||
{
|
||||
@@ -18,8 +18,7 @@ class Codec;
|
||||
typedef std::shared_ptr<Codec> PCodec;
|
||||
|
||||
class CodecMap: public std::map<int, PCodec>
|
||||
{
|
||||
};
|
||||
{};
|
||||
|
||||
class Codec
|
||||
{
|
||||
@@ -58,18 +57,28 @@ public:
|
||||
// Number of audio channels
|
||||
virtual int channels() { return 1; }
|
||||
|
||||
|
||||
// Returns size of encoded data (RTP) in bytes
|
||||
virtual int encode(const void* input, int inputBytes, void* output, int outputCapacity) = 0;
|
||||
struct EncodeResult
|
||||
{
|
||||
size_t mEncoded = 0; // Number of encoded bytes
|
||||
};
|
||||
virtual EncodeResult encode(std::span<const uint8_t> input, std::span<uint8_t> output) = 0;
|
||||
|
||||
// Returns size of decoded data (PCM signed short) in bytes
|
||||
virtual int decode(const void* input, int inputBytes, void* output, int outputCapacity) = 0;
|
||||
struct DecodeResult
|
||||
{
|
||||
size_t mDecoded = 0; // Number of decoded bytes
|
||||
bool mIsCng = false; // Should this packet to be used as CNG ? (used for AMR codecs)
|
||||
};
|
||||
virtual DecodeResult decode(std::span<const uint8_t> input, std::span<uint8_t> output) = 0;
|
||||
|
||||
// Returns size of produced data (PCM signed short) in bytes
|
||||
virtual int plc(int lostFrames, void* output, int outputCapacity) = 0;
|
||||
virtual size_t plc(int lostFrames, std::span<uint8_t> output) = 0;
|
||||
|
||||
// Returns size of codec in memory
|
||||
virtual int getSize() const { return 0; };
|
||||
virtual size_t getSize() const { return 0; };
|
||||
|
||||
virtual Audio::Format getAudioFormat() { return Audio::Format(this->samplerate(), this->channels());};
|
||||
};
|
||||
}
|
||||
#endif
|
||||
|
||||
Reference in New Issue
Block a user