- adopt RFC2833 decoder for pcap analyzer

src/engine/media/MT_AudioReceiver.cpp

@@ -201,7 +201,7 @@ RtpBuffer::FetchResult RtpBuffer::fetch()
     // See if there is enough information in buffer
     auto total = findTimelength();
 
-    while (total > mHigh && mPacketList.size() && 0ms != mHigh)
+    while (total > mHigh && mPacketList.size() > 1 && 0ms != mHigh)
     {
         ICELogMedia( << "Dropping RTP packets from jitter buffer");
         total -= mPacketList.front()->timelength();
@@ -337,7 +337,7 @@ Receiver::~Receiver()
 
 //-------------- AudioReceiver ----------------
 AudioReceiver::AudioReceiver(const CodecList::Settings& settings, MT::Statistics &stat)
-    :Receiver(stat), mBuffer(stat), mCodecSettings(settings), mCodecList(settings)
+    :Receiver(stat), mBuffer(stat), mDtmfBuffer(stat), mCodecSettings(settings), mCodecList(settings), mDtmfReceiver(stat)
 {
     // Init resamplers
     mResampler8.start(AUDIO_CHANNELS, 8000, AUDIO_SAMPLERATE);
@@ -351,6 +351,10 @@ AudioReceiver::AudioReceiver(const CodecList::Settings& settings, MT::Statistics
 
     mAvailable.setCapacity(AUDIO_SAMPLERATE * sizeof(short));
 
+    mDtmfBuffer.setPrebuffer(0ms);
+    mDtmfBuffer.setLow(0ms);
+    mDtmfBuffer.setHigh(1ms);
+
 #if defined(DUMP_DECODED)
     mDecodedDump = std::make_shared<Audio::WavFileWriter>();
     mDecodedDump->open("decoded.wav", 8000 /*G711*/, AUDIO_CHANNELS);
@@ -431,66 +435,72 @@ bool AudioReceiver::add(const std::shared_ptr<jrtplib::RTPPacket>& p, Codec** de
     // Increase codec counter
     mStat.mCodecCount[ptype]++;
 
-    // Check if codec can be handled
+    // Check if we deal with telephone-event
-    Codec* codec = nullptr;
+    if (p->GetPayloadType() == mCodecSettings.mTelephoneEvent)
-    auto codecIter = mCodecMap.find(ptype);
-    if (codecIter == mCodecMap.end())
     {
-        // Well, there is no information about the codec; skip this packet
+        *detectedCodec = nullptr;
+        mDtmfBuffer.add(p, 10ms, 8000);
     }
     else
     {
-        // Check if codec is creating lazily
+        // Look for codec
-        if (!codecIter->second)
+        // Check if codec can be handled
+        Codec* codec = nullptr;
+        auto codecIter = mCodecMap.find(ptype);
+        if (codecIter != mCodecMap.end())
         {
-            codecIter->second = mCodecList.createCodecByPayloadType(ptype);
+            // Check if codec is creating lazily
+            if (!codecIter->second)
+            {
+                codecIter->second = mCodecList.createCodecByPayloadType(ptype);
+            }
+            codec = codecIter->second.get();
+
+            // Return pointer to codec if needed.get()
+            if (detectedCodec)
+                *detectedCodec = codec;
+
+            if (mStat.mCodecName.empty() && codec)
+                mStat.mCodecName = codec->name();
+
+
+            if (!codec)
+                time_length = 10;
+            else
+            if (!codec->rtpLength())
+                time_length = codec->frameTime();
+            else
+                time_length = lround(double(payloadLength) / codec->rtpLength() * codec->frameTime());
+
+            if (codec)
+                samplerate = codec->samplerate();
         }
-        codec = codecIter->second.get();
-
-        // Return pointer to codec if needed.get()
-        if (detectedCodec)
-            *detectedCodec = codec;
-
-        if (mStat.mCodecName.empty() && codec)
-            mStat.mCodecName = codec->name();
 
+        // Process jitter anyway - can we decode payload or not
+        mJitterStats.process(p.get(), samplerate);
+        mStat.mJitter = static_cast<float>(mJitterStats.get());
 
         if (!codec)
-            time_length = 10;
+            return false; // There is no sense to add this packet into jitter buffer - we can't decode this
+
+        // Check if packet is CNG
+        if (payloadLength >= 1 && payloadLength <= 6 && (ptype == 0 || ptype == 8))
+            time_length = mLastPacketTimeLength ? mLastPacketTimeLength : 20;
         else
-        if (!codec->rtpLength())
+        // Check if packet is too short from time length side
-            time_length = codec->frameTime();
+        if (time_length < 2)
-        else
+        {
-            time_length = lround(double(payloadLength) / codec->rtpLength() * codec->frameTime());
+            // It will cause statistics to report about bad RTP packet
+            // I have to replay last packet payload here to avoid report about lost packet
+            mBuffer.add(p, std::chrono::milliseconds(time_length), samplerate);
+            return false;
+        }
 
-        if (codec)
+        // Queue packet to buffer
-            samplerate = codec->samplerate();
+        auto packet = mBuffer.add(p, std::chrono::milliseconds(time_length), samplerate).get();
+        return packet;
     }
-
+    return {};
-    // Process jitter
-    mJitterStats.process(p.get(), samplerate);
-    mStat.mJitter = static_cast<float>(mJitterStats.get());
-
-    if (!codec)
-        return false; // There is no sense to add this packet into jitter buffer - we can't decode this
-
-    // Check if packet is CNG
-    if (payloadLength >= 1 && payloadLength <= 6 && (ptype == 0 || ptype == 8))
-        time_length = mLastPacketTimeLength ? mLastPacketTimeLength : 20;
-    else
-    // Check if packet is too short from time length side
-    if (time_length < 2)
-    {
-        // It will cause statistics to report about bad RTP packet
-        // I have to replay last packet payload here to avoid report about lost packet
-        mBuffer.add(p, std::chrono::milliseconds(time_length), samplerate);
-        return false;
-    }
-
-    // Queue packet to buffer
-    auto packet = mBuffer.add(p, std::chrono::milliseconds(time_length), samplerate).get();
-
-    return packet;
 }
 
 void AudioReceiver::processDecoded(Audio::DataWindow& output, DecodeOptions options)
@@ -764,6 +774,12 @@ AudioReceiver::DecodeResult AudioReceiver::getAudioTo(Audio::DataWindow& output,
 {
     DecodeResult result = {.mStatus = DecodeResult::Status::Skip};
 
+    // Process RFC2833 here; it doesn't result in any audio - only callbacks and statistics
+    auto fr = mDtmfBuffer.fetch();
+    if (fr.mPacket && fr.mStatus == RtpBuffer::FetchResult::Status::RegularPacket)
+        mDtmfReceiver.add(fr.mPacket->rtp());
+
+
     auto produced = 0ms;
     if (mAvailable.filled() && mCodec && options.mElapsed != 0ms)
     {
@@ -968,23 +984,22 @@ DtmfReceiver::~DtmfReceiver()
 
 void DtmfReceiver::add(const std::shared_ptr<RTPPacket>& p)
 {
-    // This receiver always work in context of single RTP stream; so there is no need to put SSRC map and so on
-    if (p->GetPayloadType() != 101)
-        return;
-
     auto ev = DtmfBuilder::parseRfc2833({p->GetPayloadData(), p->GetPayloadLength()});
     if (ev.mTone != mEvent || ev.mEnd != mEventEnded)
     {
-        // New tone is here
+        if (!(mEvent == ev.mTone && !mEventEnded && ev.mEnd))
-        if (mCallback)
+        {
-            mCallback(ev.mTone);
+            // New tone is here
+            if (mCallback)
+                mCallback(ev.mTone);
 
-        // Queue statistics item
+            // Queue statistics item
-        mStat.mDtmf2833Timeline.emplace_back(Dtmf2833Event{.mTone = ev.mTone,
+            mStat.mDtmf2833Timeline.emplace_back(Dtmf2833Event{.mTone = ev.mTone,
-                                                           .mTimestamp = RtpHelper::toMicroseconds(p->GetReceiveTime())});
+                                                               .mTimestamp = RtpHelper::toMicroseconds(p->GetReceiveTime())});
 
-        // Store to avoid triggering on the packet
+            // Store to avoid triggering on the packet
-        mEvent = ev.mTone;
+            mEvent = ev.mTone;
-        mEventEnded = ev.mEnd;
+            mEventEnded = ev.mEnd;
+        }
     }
 }

src/engine/media/MT_AudioReceiver.h

@@ -136,6 +136,23 @@ protected:
     Statistics& mStat;
 };
 
+class DtmfReceiver: public Receiver
+{
+private:
+    char mEvent = 0;
+    bool mEventEnded = false;
+    std::chrono::milliseconds mEventStart = 0ms;
+    std::function<void(char)> mCallback;
+
+public:
+    DtmfReceiver(Statistics& stat);
+    ~DtmfReceiver();
+
+    void add(const std::shared_ptr<RTPPacket>& p);
+    void setCallback(std::function<void(char tone)> callback);
+};
+
+
 class AudioReceiver: public Receiver
 {
 public:
@@ -189,6 +206,9 @@ public:
 
 protected:
     RtpBuffer                           mBuffer;                // Jitter buffer itself
+    RtpBuffer                           mDtmfBuffer;            // These two (mDtmfBuffer / mDtmfReceiver) are for our analyzer stack only; in normal softphone logic DTMF packets goes via SingleAudioStream::mDtmfReceiver
+    DtmfReceiver                        mDtmfReceiver;
+
     CodecMap                            mCodecMap;
     PCodec                              mCodec;
     int                                 mFrameCount = 0;
@@ -247,21 +267,6 @@ protected:
     DecodeResult decodeEmptyTo(Audio::DataWindow& output, DecodeOptions options);
 };
 
-class DtmfReceiver: public Receiver
-{
-private:
-    char mEvent = 0;
-    bool mEventEnded = false;
-    std::chrono::milliseconds mEventStart = 0ms;
-    std::function<void(char)> mCallback;
-
-public:
-    DtmfReceiver(Statistics& stat);
-    ~DtmfReceiver();
-
-    void add(const std::shared_ptr<RTPPacket>& p);
-    void setCallback(std::function<void(char tone)> callback);
-};
 }
 
 #endif

src/engine/media/MT_CodecList.cpp

@@ -65,6 +65,8 @@ bool CodecList::Settings::contains(int ptype) const
 
     if (mGsmEfrPayloadType == ptype || mGsmFrPayloadType == ptype || mGsmHrPayloadType == ptype)
         return true;
+    if (mTelephoneEvent == ptype)
+        return true;
 
     return false;
 }
@@ -122,6 +124,9 @@ std::string CodecList::Settings::toString() const
     if (mGsmEfrPayloadType != -1)
         oss << "GSM EFR ptype: " << mGsmEfrPayloadType << " ";
 
+    if (mTelephoneEvent != -1)
+        oss << "RFC2833 DTMF ptype: " << mTelephoneEvent;
+
     for (auto& spec: mEvsSpec)
     {
         oss << "EVS ptype: " << spec.mPayloadType << ", bw: " << spec.mBandwidth << ", enc: " << (spec.mEncodingType == EvsSpec::Encoding_MIME ? "mime" : "g192") << " ";
@@ -132,6 +137,7 @@ std::string CodecList::Settings::toString() const
         oss << "OPUS ptype: " << spec.mPayloadType << ", rate: " << spec.mRate << ", channels: " << spec.mChannels << std::endl;
     }
 
+
     return oss.str();
 }
 
@@ -151,6 +157,7 @@ void CodecList::Settings::clear()
     mGsmEfrPayloadType = -1;
     mGsmFrPayloadType = -1;
     mGsmHrPayloadType = -1;
+    mTelephoneEvent = -1;
 }
 
 bool CodecList::Settings::EvsSpec::isValid() const
@@ -268,15 +275,16 @@ CodecList::Settings CodecList::Settings::parseSdp(const std::list<resip::Codec>&
             }
         } else if (codec_name == "EVS") {
             r.mEvsSpec.push_back({ptype});
-        }
+        } else if (codec_name == "TELEPHONE-EVENT")
+            r.mTelephoneEvent = ptype;
     }
     return r;
 }
 
 bool CodecList::Settings::operator == (const Settings& rhs) const
 {
-    if (std::tie(mWrapIuUP, mSkipDecode, mIsac16KPayloadType, mIsac32KPayloadType, mIlbc20PayloadType, mIlbc30PayloadType, mGsmFrPayloadType, mGsmFrPayloadLength, mGsmEfrPayloadType, mGsmHrPayloadType) !=
+    if (std::tie(mWrapIuUP, mSkipDecode, mIsac16KPayloadType, mIsac32KPayloadType, mIlbc20PayloadType, mIlbc30PayloadType, mGsmFrPayloadType, mGsmFrPayloadLength, mGsmEfrPayloadType, mGsmHrPayloadType, mTelephoneEvent) !=
-        std::tie(rhs.mWrapIuUP, rhs.mSkipDecode, rhs.mIsac16KPayloadType, rhs.mIsac32KPayloadType, rhs.mIlbc20PayloadType, rhs.mIlbc30PayloadType, rhs.mGsmFrPayloadType, rhs.mGsmFrPayloadLength, rhs.mGsmEfrPayloadType, rhs.mGsmHrPayloadType))
+        std::tie(rhs.mWrapIuUP, rhs.mSkipDecode, rhs.mIsac16KPayloadType, rhs.mIsac32KPayloadType, rhs.mIlbc20PayloadType, rhs.mIlbc30PayloadType, rhs.mGsmFrPayloadType, rhs.mGsmFrPayloadLength, rhs.mGsmEfrPayloadType, rhs.mGsmHrPayloadType, rhs.mTelephoneEvent))
         return false;
 
     if (mAmrNbOctetPayloadType != rhs.mAmrNbOctetPayloadType)
@@ -306,6 +314,9 @@ bool CodecList::Settings::operator == (const Settings& rhs) const
         if (mOpusSpec[i] != rhs.mOpusSpec[i])
             return false;
 
+    if (mTelephoneEvent != rhs.mTelephoneEvent)
+        return false;
+
     return true;
 }
 

src/engine/media/MT_CodecList.h

@@ -28,6 +28,9 @@ public:
         bool mWrapIuUP              = false;
         bool mSkipDecode            = false;
 
+        // RFC2833 DTMF
+        int mTelephoneEvent = -1;
+
         // AMR payload types
         std::set<int64_t> mAmrWbPayloadType       = { };
         std::set<int64_t> mAmrNbPayloadType       = { };

src/engine/media/MT_SingleAudioStream.cpp

@@ -23,7 +23,7 @@ SingleAudioStream::~SingleAudioStream()
 void SingleAudioStream::process(const std::shared_ptr<jrtplib::RTPPacket>& packet)
 {
     ICELogMedia(<< "Processing incoming RTP/RTCP packet");
-    if (packet->GetPayloadType() == 101/*resip::Codec::TelephoneEvent.payloadType()*/)
+    if (packet->GetPayloadType() == mReceiver.getCodecSettings().mTelephoneEvent)
         mDtmfReceiver.add(packet);
     else
         mReceiver.add(packet);