added stdev method of jitter calc to InboundAudioStream

assignment-client/src/audio/AudioMixerClientData.cpp

@@ -205,7 +205,7 @@ QString AudioMixerClientData::getAudioStreamStatsString() const {
     if (avatarRingBuffer) {
         AudioStreamStats streamStats = avatarRingBuffer->getAudioStreamStats();
         result += " UPSTREAM.mic.desired:" + QString::number(streamStats._ringBufferDesiredJitterBufferFrames)
-            + " desired_calc:" + QString::number(avatarRingBuffer->getCalculatedDesiredJitterBufferFrames())
+            + " desired_calc:" + QString::number(avatarRingBuffer->getCalculatedJitterBufferFrames())
             + " available_avg_10s:" + QString::number(streamStats._ringBufferFramesAvailableAverage)
             + " available:" + QString::number(streamStats._ringBufferFramesAvailable)
             + " starves:" + QString::number(streamStats._ringBufferStarveCount)
@@ -225,11 +225,11 @@ QString AudioMixerClientData::getAudioStreamStatsString() const {
     }
     
     QHash<QUuid, PositionalAudioRingBuffer*>::ConstIterator i;
-    for (i = _ringBuffers.constBegin(); i != _ringBuffers.constEnd; i++) {
+    for (i = _ringBuffers.constBegin(); i != _ringBuffers.constEnd(); i++) {
         if (i.value()->getType() == PositionalAudioRingBuffer::Injector) {
             AudioStreamStats streamStats = i.value()->getAudioStreamStats();
             result += " UPSTREAM.inj.desired:" + QString::number(streamStats._ringBufferDesiredJitterBufferFrames)
-                + " desired_calc:" + QString::number(i.value()->getCalculatedDesiredJitterBufferFrames())
+                + " desired_calc:" + QString::number(i.value()->getCalculatedJitterBufferFrames())
                 + " available_avg_10s:" + QString::number(streamStats._ringBufferFramesAvailableAverage)
                 + " available:" + QString::number(streamStats._ringBufferFramesAvailable)
                 + " starves:" + QString::number(streamStats._ringBufferStarveCount)

interface/src/Audio.cpp

@@ -76,9 +76,9 @@ Audio::Audio(int16_t initialJitterBufferSamples, QObject* parent) :
     // this delay will slowly add up and the longer someone runs, they more delayed their audio will be.
     _inputRingBuffer(0),
 #ifdef _WIN32
-    _ringBuffer(NETWORK_BUFFER_LENGTH_SAMPLES_STEREO, 100, true),
+    _ringBuffer(NETWORK_BUFFER_LENGTH_SAMPLES_STEREO, 100, true, true),
 #else
-    _ringBuffer(NETWORK_BUFFER_LENGTH_SAMPLES_STEREO, 10, true), // DO NOT CHANGE THIS UNLESS YOU SOLVE THE AUDIO DEVICE DRIFT PROBLEM!!!
+    _ringBuffer(NETWORK_BUFFER_LENGTH_SAMPLES_STEREO, 10, true, true), // DO NOT CHANGE THIS UNLESS YOU SOLVE THE AUDIO DEVICE DRIFT PROBLEM!!!
 #endif  
     _isStereoInput(false),
     _averagedLatency(0.0),

libraries/audio/src/InboundAudioStream.cpp

@@ -12,11 +12,14 @@
 #include "InboundAudioStream.h"
 #include "PacketHeaders.h"
 
-InboundAudioStream::InboundAudioStream(int numFrameSamples, int numFramesCapacity, bool dynamicJitterBuffers) :
+InboundAudioStream::InboundAudioStream(int numFrameSamples, int numFramesCapacity, bool dynamicJitterBuffers, bool useStDevForJitterCalc) :
     _ringBuffer(numFrameSamples, false, numFramesCapacity),
     _lastPopSucceeded(false),
     _lastPopOutput(),
     _dynamicJitterBuffers(dynamicJitterBuffers),
+    _useStDevForJitterCalc(useStDevForJitterCalc),
+    _calculatedJitterBufferFramesUsingMaxGap(0),
+    _calculatedJitterBufferFramesUsingStDev(0),
     _desiredJitterBufferFrames(1),
     _isStarved(true),
     _hasStarted(false),
@@ -143,46 +146,51 @@ void InboundAudioStream::starved() {
     _starveCount++;
 }
 
-
+int InboundAudioStream::clampDesiredJitterBufferFramesValue(int desired) const {
-int InboundAudioStream::getCalculatedDesiredJitterBufferFrames() const {
+    const int MIN_FRAMES_DESIRED = 0;
-    const float USECS_PER_FRAME = NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL * USECS_PER_SECOND / (float)SAMPLE_RATE;
+    const int MAX_FRAMES_DESIRED = _ringBuffer.getFrameCapacity();
-
+    return glm::clamp(desired, MIN_FRAMES_DESIRED, MAX_FRAMES_DESIRED);
-    int calculatedDesiredJitterBufferFrames = ceilf((float)_interframeTimeGapStatsForJitterCalc.getWindowMax() / USECS_PER_FRAME);
-    if (calculatedDesiredJitterBufferFrames < 1) {
-        calculatedDesiredJitterBufferFrames = 1;
-    }
-    return calculatedDesiredJitterBufferFrames;
 }
 
-
 SequenceNumberStats::ArrivalInfo InboundAudioStream::frameReceivedUpdateNetworkStats(quint16 sequenceNumber, const QUuid& senderUUID) {
-    const int NUM_INITIAL_PACKETS_DISCARD = 3;
+    // track the sequence number we received
-
     SequenceNumberStats::ArrivalInfo arrivalInfo = _incomingSequenceNumberStats.sequenceNumberReceived(sequenceNumber, senderUUID);
 
-    // update the two time gap stats we're keeping
+    // update our timegap stats and desired jitter buffer frames if necessary
+    // discard the first few packets we receive since they usually have gaps that aren't represensative of normal jitter
+    const int NUM_INITIAL_PACKETS_DISCARD = 3;
     quint64 now = usecTimestampNow();
-    if (_incomingSequenceNumberStats.getNumReceived() >= NUM_INITIAL_PACKETS_DISCARD) {
+    if (_incomingSequenceNumberStats.getNumReceived() > NUM_INITIAL_PACKETS_DISCARD) {
         quint64 gap = now - _lastFrameReceivedTime;
-        _interframeTimeGapStatsForJitterCalc.update(gap);
         _interframeTimeGapStatsForStatsPacket.update(gap);
-    }
-    _lastFrameReceivedTime = now;
 
-    // recalculate the _desiredJitterBufferFrames if _interframeTimeGapStatsForJitterCalc has updated stats for us
+        const float USECS_PER_FRAME = NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL * USECS_PER_SECOND / (float)SAMPLE_RATE;
-    if (_interframeTimeGapStatsForJitterCalc.getNewStatsAvailableFlag()) {
-        if (!_dynamicJitterBuffers) {
-            _desiredJitterBufferFrames = 1; // HACK to see if this fixes the audio silence
-        } else {
-            _desiredJitterBufferFrames = getCalculatedDesiredJitterBufferFrames();
 
-            const int maxDesired = _ringBuffer.getFrameCapacity() - 1;
+        // update stats for Freddy's method of jitter calc
-            if (_desiredJitterBufferFrames > maxDesired) {
+        _interframeTimeGapStatsForJitterCalc.update(gap);
-                _desiredJitterBufferFrames = maxDesired;
+        if (_interframeTimeGapStatsForJitterCalc.getNewStatsAvailableFlag()) {
+            _calculatedJitterBufferFramesUsingMaxGap = ceilf((float)_interframeTimeGapStatsForJitterCalc.getWindowMax() / USECS_PER_FRAME);
+            _interframeTimeGapStatsForJitterCalc.clearNewStatsAvailableFlag();
+
+            if (_dynamicJitterBuffers && !_useStDevForJitterCalc) {
+                _desiredJitterBufferFrames = clampDesiredJitterBufferFramesValue(_calculatedJitterBufferFramesUsingMaxGap);
+            }
+        }
+
+        // update stats for Philip's method of jitter calc
+        _stdev.addValue(gap);
+        const int STANDARD_DEVIATION_SAMPLE_COUNT = 500;
+        if (_stdev.getSamples() > STANDARD_DEVIATION_SAMPLE_COUNT) {
+            const float NUM_STANDARD_DEVIATIONS = 3.0f;
+            _calculatedJitterBufferFramesUsingStDev = (int)ceilf(2 * (NUM_STANDARD_DEVIATIONS * _stdev.getStDev()) / USECS_PER_FRAME) + 1;
+            _stdev.reset();
+
+            if (_dynamicJitterBuffers && _useStDevForJitterCalc) {
+                _desiredJitterBufferFrames = clampDesiredJitterBufferFramesValue(_calculatedJitterBufferFramesUsingStDev);
             }
         }
-        _interframeTimeGapStatsForJitterCalc.clearNewStatsAvailableFlag();
     }
+    _lastFrameReceivedTime = now;
 
     return arrivalInfo;
 }

libraries/audio/src/InboundAudioStream.h

@@ -18,6 +18,7 @@
 #include "SequenceNumberStats.h"
 #include "AudioStreamStats.h"
 #include "PacketHeaders.h"
+#include "StdDev.h"
 
 // the time gaps stats for _desiredJitterBufferFrames calculation
 // will recalculate the max for the past 5000 samples every 500 samples
@@ -44,7 +45,7 @@ const int INBOUND_RING_BUFFER_FRAME_CAPACITY = 100;
 class InboundAudioStream : public NodeData {
     Q_OBJECT
 public:
-    InboundAudioStream(int numFrameSamples, int numFramesCapacity, bool dynamicJitterBuffers);
+    InboundAudioStream(int numFrameSamples, int numFramesCapacity, bool dynamicJitterBuffers, bool useStDevForJitterCalc = false);
 
     void reset();
     void resetStats();
@@ -67,7 +68,15 @@ public:
 
     virtual AudioStreamStats getAudioStreamStats() const;
 
-    int getCalculatedDesiredJitterBufferFrames() const;
+    /// returns the desired number of jitter buffer frames under the dyanmic jitter buffers scheme
+    int getCalculatedJitterBufferFrames() const { return _useStDevForJitterCalc ? 
+        _calculatedJitterBufferFramesUsingStDev : _calculatedJitterBufferFramesUsingMaxGap; };
+
+    /// returns the desired number of jitter buffer frames using Philip's method
+    int getCalculatedJitterBufferFramesUsingStDev() const { return _calculatedJitterBufferFramesUsingStDev; }
+
+    /// returns the desired number of jitter buffer frames using Freddy's method
+    int getCalculatedJitterBufferFramesUsingMaxGap() const { return _calculatedJitterBufferFramesUsingMaxGap; }
 
     int getDesiredJitterBufferFrames() const { return _desiredJitterBufferFrames; }
     int getNumFrameSamples() const { return _ringBuffer.getNumFrameSamples(); }
@@ -86,6 +95,8 @@ public:
 private:
     void starved();
 
+    int clampDesiredJitterBufferFramesValue(int desired) const;
+
 protected:
     // disallow copying of InboundAudioStream objects
     InboundAudioStream(const InboundAudioStream&);
@@ -110,6 +121,10 @@ protected:
     AudioRingBuffer::ConstIterator _lastPopOutput;
 
     bool _dynamicJitterBuffers;
+    bool _useStDevForJitterCalc;
+    
+    int _calculatedJitterBufferFramesUsingMaxGap;
+    int _calculatedJitterBufferFramesUsingStDev;
     int _desiredJitterBufferFrames;
 
     bool _isStarved;
@@ -125,6 +140,7 @@ protected:
 
     quint64 _lastFrameReceivedTime;
     MovingMinMaxAvg<quint64> _interframeTimeGapStatsForJitterCalc;
+    StDev _stdev;
     MovingMinMaxAvg<quint64> _interframeTimeGapStatsForStatsPacket;
     
     // TODO: change this to time-weighted moving avg

libraries/audio/src/InboundMixedAudioStream.cpp

@@ -1,8 +1,8 @@
 
 #include "InboundMixedAudioStream.h"
 
-InboundMixedAudioStream::InboundMixedAudioStream(int numFrameSamples, int numFramesCapacity, bool dynamicJitterBuffers)
+InboundMixedAudioStream::InboundMixedAudioStream(int numFrameSamples, int numFramesCapacity, bool dynamicJitterBuffers, bool useStDevForJitterCalc)
-    : InboundAudioStream(numFrameSamples, numFramesCapacity, dynamicJitterBuffers)
+    : InboundAudioStream(numFrameSamples, numFramesCapacity, dynamicJitterBuffers, useStDevForJitterCalc)
 {
 }
 

libraries/audio/src/InboundMixedAudioStream.h

@@ -4,7 +4,7 @@
 
 class InboundMixedAudioStream : public InboundAudioStream {
 public:
-    InboundMixedAudioStream(int numFrameSamples, int numFramesCapacity, bool dynamicJitterBuffers);
+    InboundMixedAudioStream(int numFrameSamples, int numFramesCapacity, bool dynamicJitterBuffers, bool useStDevForJitterCalc = false);
 
     float getNextOutputFrameLoudness() const { return _ringBuffer.getNextOutputFrameLoudness(); }
 

libraries/shared/src/StdDev.cpp

@@ -29,7 +29,7 @@ void StDev::addValue(float v) {
     if (sampleCount == MAX_STDEV_SAMPLES) sampleCount = 0;
 }
 
-float StDev::getAverage() {
+float StDev::getAverage() const {
     float average = 0;
     for (int i = 0; i < sampleCount; i++) {
         average += data[i];
@@ -49,7 +49,7 @@ float StDev::getMax() {
     else return 0;
 }*/
 
-float StDev::getStDev() {
+float StDev::getStDev() const {
     float average = getAverage();
     float stdev = 0;
     for (int i = 0; i < sampleCount; i++) {

libraries/shared/src/StdDev.h

@@ -17,8 +17,8 @@ class StDev {
         StDev();
         void reset();
         void addValue(float v);
-        float getAverage();
+        float getAverage() const;
-        float getStDev();
+        float getStDev() const;
         int getSamples() const { return sampleCount; }
     private:
         float * data;