mirror of
https://github.com/lubosz/overte.git
synced 2025-04-23 01:04:06 +02:00
moved time gap history to PositionalAudioRingBuffer, untested
removed hard-coded jitter buffer length.
This commit is contained in:
parent
100bc022ec
commit
435b5094a0
9 changed files with 118 additions and 104 deletions
|
@ -54,9 +54,6 @@
|
|||
|
||||
#include "AudioMixer.h"
|
||||
|
||||
const short JITTER_BUFFER_MSECS = 12;
|
||||
const short JITTER_BUFFER_SAMPLES = JITTER_BUFFER_MSECS * (SAMPLE_RATE / 1000.0);
|
||||
|
||||
const float LOUDNESS_TO_DISTANCE_RATIO = 0.00001f;
|
||||
|
||||
const QString AUDIO_MIXER_LOGGING_TARGET_NAME = "audio-mixer";
|
||||
|
@ -487,8 +484,7 @@ void AudioMixer::run() {
|
|||
|
||||
foreach (const SharedNodePointer& node, nodeList->getNodeHash()) {
|
||||
if (node->getLinkedData()) {
|
||||
((AudioMixerClientData*) node->getLinkedData())->checkBuffersBeforeFrameSend(JITTER_BUFFER_SAMPLES,
|
||||
_sourceUnattenuatedZone,
|
||||
((AudioMixerClientData*) node->getLinkedData())->checkBuffersBeforeFrameSend(_sourceUnattenuatedZone,
|
||||
_listenerUnattenuatedZone);
|
||||
}
|
||||
}
|
||||
|
|
|
@ -98,10 +98,9 @@ int AudioMixerClientData::parseData(const QByteArray& packet) {
|
|||
return 0;
|
||||
}
|
||||
|
||||
void AudioMixerClientData::checkBuffersBeforeFrameSend(int jitterBufferLengthSamples,
|
||||
AABox* checkSourceZone, AABox* listenerZone) {
|
||||
void AudioMixerClientData::checkBuffersBeforeFrameSend(AABox* checkSourceZone, AABox* listenerZone) {
|
||||
for (int i = 0; i < _ringBuffers.size(); i++) {
|
||||
if (_ringBuffers[i]->shouldBeAddedToMix(jitterBufferLengthSamples)) {
|
||||
if (_ringBuffers[i]->shouldBeAddedToMix()) {
|
||||
// this is a ring buffer that is ready to go
|
||||
// set its flag so we know to push its buffer when all is said and done
|
||||
_ringBuffers[i]->setWillBeAddedToMix(true);
|
||||
|
|
|
@ -27,8 +27,7 @@ public:
|
|||
AvatarAudioRingBuffer* getAvatarAudioRingBuffer() const;
|
||||
|
||||
int parseData(const QByteArray& packet);
|
||||
void checkBuffersBeforeFrameSend(int jitterBufferLengthSamples,
|
||||
AABox* checkSourceZone = NULL, AABox* listenerZone = NULL);
|
||||
void checkBuffersBeforeFrameSend(AABox* checkSourceZone = NULL, AABox* listenerZone = NULL);
|
||||
void pushBuffersAfterFrameSend();
|
||||
private:
|
||||
QList<PositionalAudioRingBuffer*> _ringBuffers;
|
||||
|
|
|
@ -19,7 +19,8 @@ AvatarAudioRingBuffer::AvatarAudioRingBuffer(bool isStereo) :
|
|||
}
|
||||
|
||||
int AvatarAudioRingBuffer::parseData(const QByteArray& packet) {
|
||||
_timeGapHistory.frameReceived();
|
||||
_interframeTimeGapHistory.frameReceived();
|
||||
updateDesiredJitterBufferNumSamples();
|
||||
|
||||
_shouldLoopbackForNode = (packetTypeForPacket(packet) == PacketTypeMicrophoneAudioWithEcho);
|
||||
return PositionalAudioRingBuffer::parseData(packet);
|
||||
|
|
|
@ -12,7 +12,6 @@
|
|||
#include <cstring>
|
||||
#include <functional>
|
||||
#include <math.h>
|
||||
#include "SharedUtil.h"
|
||||
|
||||
#include <QtCore/QDebug>
|
||||
|
||||
|
@ -20,62 +19,6 @@
|
|||
|
||||
#include "AudioRingBuffer.h"
|
||||
|
||||
InterframeTimeGapHistory::InterframeTimeGapHistory()
|
||||
: _lastFrameReceivedTime(0),
|
||||
_numSamplesInCurrentInterval(0),
|
||||
_currentIntervalMaxGap(0),
|
||||
_newestIntervalMaxGapAt(0),
|
||||
_windowMaxGap(0),
|
||||
_newWindowMaxGapAvailable(false)
|
||||
{
|
||||
memset(_intervalMaxGaps, 0, TIME_GAP_NUM_INTERVALS_IN_WINDOW*sizeof(quint64));
|
||||
}
|
||||
|
||||
void InterframeTimeGapHistory::frameReceived() {
|
||||
quint64 now = usecTimestampNow();
|
||||
|
||||
// make sure this isn't the first time frameReceived() is called, meaning there's actually a gap to calculate.
|
||||
if (_lastFrameReceivedTime != 0) {
|
||||
quint64 gap = now - _lastFrameReceivedTime;
|
||||
|
||||
// update the current interval max
|
||||
if (gap > _currentIntervalMaxGap) {
|
||||
_currentIntervalMaxGap = gap;
|
||||
}
|
||||
_numSamplesInCurrentInterval++;
|
||||
|
||||
// if the current interval of samples is now full, record it in our interval maxes
|
||||
if (_numSamplesInCurrentInterval == TIME_GAP_NUM_SAMPLES_IN_INTERVAL) {
|
||||
|
||||
// find location to insert this interval's max (increment index cyclically)
|
||||
_newestIntervalMaxGapAt = _newestIntervalMaxGapAt == TIME_GAP_NUM_INTERVALS_IN_WINDOW - 1 ? 0 : _newestIntervalMaxGapAt + 1;
|
||||
|
||||
// record the current interval's max gap as the newest
|
||||
_intervalMaxGaps[_newestIntervalMaxGapAt] = _currentIntervalMaxGap;
|
||||
|
||||
// update the window max gap, which is the max out of all the past intervals' max gaps
|
||||
_windowMaxGap = 0;
|
||||
for (int i = 0; i < TIME_GAP_NUM_INTERVALS_IN_WINDOW; i++) {
|
||||
if (_intervalMaxGaps[i] > _windowMaxGap) {
|
||||
_windowMaxGap = _intervalMaxGaps[i];
|
||||
}
|
||||
}
|
||||
_newWindowMaxGapAvailable = true;
|
||||
|
||||
// reset the current interval
|
||||
_numSamplesInCurrentInterval = 0;
|
||||
_currentIntervalMaxGap = 0;
|
||||
}
|
||||
}
|
||||
_lastFrameReceivedTime = now;
|
||||
}
|
||||
|
||||
quint64 InterframeTimeGapHistory::getPastWindowMaxGap() {
|
||||
_newWindowMaxGapAvailable = false;
|
||||
return _windowMaxGap;
|
||||
}
|
||||
|
||||
|
||||
AudioRingBuffer::AudioRingBuffer(int numFrameSamples, bool randomAccessMode) :
|
||||
NodeData(),
|
||||
_sampleCapacity(numFrameSamples * RING_BUFFER_LENGTH_FRAMES),
|
||||
|
|
|
@ -21,31 +21,6 @@
|
|||
|
||||
#include "NodeData.h"
|
||||
|
||||
// this means that every 500 samples, the max for the past 10*500 samples will be calculated
|
||||
const int TIME_GAP_NUM_SAMPLES_IN_INTERVAL = 500;
|
||||
const int TIME_GAP_NUM_INTERVALS_IN_WINDOW = 10;
|
||||
|
||||
// class used to track time between incoming frames for the purpose of varying the jitter buffer length
|
||||
class InterframeTimeGapHistory {
|
||||
public:
|
||||
InterframeTimeGapHistory();
|
||||
|
||||
void frameReceived();
|
||||
bool isNewWindowMaxGapAvailable() const { return _newWindowMaxGapAvailable; }
|
||||
quint64 getPastWindowMaxGap();
|
||||
|
||||
private:
|
||||
quint64 _lastFrameReceivedTime;
|
||||
|
||||
int _numSamplesInCurrentInterval;
|
||||
quint64 _currentIntervalMaxGap;
|
||||
quint64 _intervalMaxGaps[TIME_GAP_NUM_INTERVALS_IN_WINDOW];
|
||||
int _newestIntervalMaxGapAt;
|
||||
quint64 _windowMaxGap;
|
||||
bool _newWindowMaxGapAvailable;
|
||||
};
|
||||
|
||||
|
||||
const int SAMPLE_RATE = 24000;
|
||||
|
||||
const int NETWORK_BUFFER_LENGTH_BYTES_STEREO = 1024;
|
||||
|
@ -99,15 +74,14 @@ public:
|
|||
bool hasStarted() const { return _hasStarted; }
|
||||
|
||||
void addSilentFrame(int numSilentSamples);
|
||||
|
||||
InterframeTimeGapHistory& getInterframeTimeGapHistory() { return _timeGapHistory; }
|
||||
|
||||
protected:
|
||||
// disallow copying of AudioRingBuffer objects
|
||||
AudioRingBuffer(const AudioRingBuffer&);
|
||||
AudioRingBuffer& operator= (const AudioRingBuffer&);
|
||||
|
||||
int16_t* shiftedPositionAccomodatingWrap(int16_t* position, int numSamplesShift) const;
|
||||
|
||||
|
||||
int _sampleCapacity;
|
||||
int _numFrameSamples;
|
||||
int16_t* _nextOutput;
|
||||
|
@ -116,8 +90,6 @@ protected:
|
|||
bool _isStarved;
|
||||
bool _hasStarted;
|
||||
bool _randomAccessMode; /// will this ringbuffer be used for random access? if so, do some special processing
|
||||
|
||||
InterframeTimeGapHistory _timeGapHistory;
|
||||
};
|
||||
|
||||
#endif // hifi_AudioRingBuffer_h
|
||||
|
|
|
@ -31,7 +31,8 @@ InjectedAudioRingBuffer::InjectedAudioRingBuffer(const QUuid& streamIdentifier)
|
|||
const uchar MAX_INJECTOR_VOLUME = 255;
|
||||
|
||||
int InjectedAudioRingBuffer::parseData(const QByteArray& packet) {
|
||||
_timeGapHistory.frameReceived();
|
||||
_interframeTimeGapHistory.frameReceived();
|
||||
updateDesiredJitterBufferNumSamples();
|
||||
|
||||
// setup a data stream to read from this packet
|
||||
QDataStream packetStream(packet);
|
||||
|
|
|
@ -19,6 +19,63 @@
|
|||
#include <UUID.h>
|
||||
|
||||
#include "PositionalAudioRingBuffer.h"
|
||||
#include "SharedUtil.h"
|
||||
|
||||
InterframeTimeGapHistory::InterframeTimeGapHistory()
|
||||
: _lastFrameReceivedTime(0),
|
||||
_numSamplesInCurrentInterval(0),
|
||||
_currentIntervalMaxGap(0),
|
||||
_newestIntervalMaxGapAt(0),
|
||||
_windowMaxGap(0),
|
||||
_newWindowMaxGapAvailable(false)
|
||||
{
|
||||
memset(_intervalMaxGaps, 0, TIME_GAP_NUM_INTERVALS_IN_WINDOW*sizeof(quint64));
|
||||
}
|
||||
|
||||
void InterframeTimeGapHistory::frameReceived() {
|
||||
quint64 now = usecTimestampNow();
|
||||
|
||||
// make sure this isn't the first time frameReceived() is called, meaning there's actually a gap to calculate.
|
||||
if (_lastFrameReceivedTime != 0) {
|
||||
quint64 gap = now - _lastFrameReceivedTime;
|
||||
|
||||
// update the current interval max
|
||||
if (gap > _currentIntervalMaxGap) {
|
||||
_currentIntervalMaxGap = gap;
|
||||
}
|
||||
_numSamplesInCurrentInterval++;
|
||||
|
||||
// if the current interval of samples is now full, record it in our interval maxes
|
||||
if (_numSamplesInCurrentInterval == TIME_GAP_NUM_SAMPLES_IN_INTERVAL) {
|
||||
|
||||
// find location to insert this interval's max (increment index cyclically)
|
||||
_newestIntervalMaxGapAt = _newestIntervalMaxGapAt == TIME_GAP_NUM_INTERVALS_IN_WINDOW - 1 ? 0 : _newestIntervalMaxGapAt + 1;
|
||||
|
||||
// record the current interval's max gap as the newest
|
||||
_intervalMaxGaps[_newestIntervalMaxGapAt] = _currentIntervalMaxGap;
|
||||
|
||||
// update the window max gap, which is the max out of all the past intervals' max gaps
|
||||
_windowMaxGap = 0;
|
||||
for (int i = 0; i < TIME_GAP_NUM_INTERVALS_IN_WINDOW; i++) {
|
||||
if (_intervalMaxGaps[i] > _windowMaxGap) {
|
||||
_windowMaxGap = _intervalMaxGaps[i];
|
||||
}
|
||||
}
|
||||
_newWindowMaxGapAvailable = true;
|
||||
|
||||
// reset the current interval
|
||||
_numSamplesInCurrentInterval = 0;
|
||||
_currentIntervalMaxGap = 0;
|
||||
}
|
||||
}
|
||||
_lastFrameReceivedTime = now;
|
||||
}
|
||||
|
||||
quint64 InterframeTimeGapHistory::getWindowMaxGap() {
|
||||
_newWindowMaxGapAvailable = false;
|
||||
return _windowMaxGap;
|
||||
}
|
||||
|
||||
|
||||
PositionalAudioRingBuffer::PositionalAudioRingBuffer(PositionalAudioRingBuffer::Type type, bool isStereo) :
|
||||
AudioRingBuffer(isStereo ? NETWORK_BUFFER_LENGTH_SAMPLES_STEREO : NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL),
|
||||
|
@ -29,7 +86,8 @@ PositionalAudioRingBuffer::PositionalAudioRingBuffer(PositionalAudioRingBuffer::
|
|||
_shouldLoopbackForNode(false),
|
||||
_shouldOutputStarveDebug(true),
|
||||
_isStereo(isStereo),
|
||||
_listenerUnattenuatedZone(NULL)
|
||||
_listenerUnattenuatedZone(NULL),
|
||||
_desiredJitterBufferNumSamples(getNumSamplesPerFrame())
|
||||
{
|
||||
|
||||
}
|
||||
|
@ -106,14 +164,17 @@ void PositionalAudioRingBuffer::updateNextOutputTrailingLoudness() {
|
|||
}
|
||||
}
|
||||
|
||||
bool PositionalAudioRingBuffer::shouldBeAddedToMix(int numJitterBufferSamples) {
|
||||
if (!isNotStarvedOrHasMinimumSamples(NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL + numJitterBufferSamples)) {
|
||||
bool PositionalAudioRingBuffer::shouldBeAddedToMix() {
|
||||
|
||||
int samplesPerFrame = getNumSamplesPerFrame();
|
||||
|
||||
if (!isNotStarvedOrHasMinimumSamples(samplesPerFrame + _desiredJitterBufferNumSamples)) {
|
||||
if (_shouldOutputStarveDebug) {
|
||||
_shouldOutputStarveDebug = false;
|
||||
}
|
||||
|
||||
return false;
|
||||
} else if (samplesAvailable() < NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL) {
|
||||
} else if (samplesAvailable() < samplesPerFrame) {
|
||||
_isStarved = true;
|
||||
|
||||
// reset our _shouldOutputStarveDebug to true so the next is printed
|
||||
|
@ -132,3 +193,13 @@ bool PositionalAudioRingBuffer::shouldBeAddedToMix(int numJitterBufferSamples) {
|
|||
|
||||
return false;
|
||||
}
|
||||
|
||||
void PositionalAudioRingBuffer::updateDesiredJitterBufferNumSamples() {
|
||||
|
||||
const float USECS_PER_FRAME = NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL * USECS_PER_SECOND / (float)SAMPLE_RATE;
|
||||
|
||||
if (_interframeTimeGapHistory.hasNewWindowMaxGapAvailable()) {
|
||||
int desiredJitterBufferNumFrames = (int)((float)_interframeTimeGapHistory.getWindowMaxGap() / USECS_PER_FRAME + 0.5f);
|
||||
_desiredJitterBufferNumSamples = desiredJitterBufferNumFrames * getNumSamplesPerFrame();
|
||||
}
|
||||
}
|
|
@ -18,6 +18,31 @@
|
|||
|
||||
#include "AudioRingBuffer.h"
|
||||
|
||||
// this means that every 500 samples, the max for the past 10*500 samples will be calculated
|
||||
const int TIME_GAP_NUM_SAMPLES_IN_INTERVAL = 500;
|
||||
const int TIME_GAP_NUM_INTERVALS_IN_WINDOW = 10;
|
||||
|
||||
// class used to track time between incoming frames for the purpose of varying the jitter buffer length
|
||||
class InterframeTimeGapHistory {
|
||||
public:
|
||||
InterframeTimeGapHistory();
|
||||
|
||||
void frameReceived();
|
||||
bool hasNewWindowMaxGapAvailable() const { return _newWindowMaxGapAvailable; }
|
||||
quint64 getWindowMaxGap();
|
||||
|
||||
private:
|
||||
quint64 _lastFrameReceivedTime;
|
||||
|
||||
int _numSamplesInCurrentInterval;
|
||||
quint64 _currentIntervalMaxGap;
|
||||
quint64 _intervalMaxGaps[TIME_GAP_NUM_INTERVALS_IN_WINDOW];
|
||||
int _newestIntervalMaxGapAt;
|
||||
quint64 _windowMaxGap;
|
||||
bool _newWindowMaxGapAvailable;
|
||||
};
|
||||
|
||||
|
||||
class PositionalAudioRingBuffer : public AudioRingBuffer {
|
||||
public:
|
||||
enum Type {
|
||||
|
@ -34,7 +59,7 @@ public:
|
|||
void updateNextOutputTrailingLoudness();
|
||||
float getNextOutputTrailingLoudness() const { return _nextOutputTrailingLoudness; }
|
||||
|
||||
bool shouldBeAddedToMix(int numJitterBufferSamples);
|
||||
bool shouldBeAddedToMix();
|
||||
|
||||
bool willBeAddedToMix() const { return _willBeAddedToMix; }
|
||||
void setWillBeAddedToMix(bool willBeAddedToMix) { _willBeAddedToMix = willBeAddedToMix; }
|
||||
|
@ -50,10 +75,14 @@ public:
|
|||
AABox* getListenerUnattenuatedZone() const { return _listenerUnattenuatedZone; }
|
||||
void setListenerUnattenuatedZone(AABox* listenerUnattenuatedZone) { _listenerUnattenuatedZone = listenerUnattenuatedZone; }
|
||||
|
||||
int getNumSamplesPerFrame() const { return _isStereo ? NETWORK_BUFFER_LENGTH_SAMPLES_STEREO : NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL; }
|
||||
|
||||
protected:
|
||||
// disallow copying of PositionalAudioRingBuffer objects
|
||||
PositionalAudioRingBuffer(const PositionalAudioRingBuffer&);
|
||||
PositionalAudioRingBuffer& operator= (const PositionalAudioRingBuffer&);
|
||||
|
||||
void updateDesiredJitterBufferNumSamples();
|
||||
|
||||
PositionalAudioRingBuffer::Type _type;
|
||||
glm::vec3 _position;
|
||||
|
@ -65,6 +94,9 @@ protected:
|
|||
|
||||
float _nextOutputTrailingLoudness;
|
||||
AABox* _listenerUnattenuatedZone;
|
||||
|
||||
InterframeTimeGapHistory _interframeTimeGapHistory;
|
||||
int _desiredJitterBufferNumSamples;
|
||||
};
|
||||
|
||||
#endif // hifi_PositionalAudioRingBuffer_h
|
||||
|
|
Loading…
Reference in a new issue