Initial cut of htrf for mono localOnly injectors

Probably need to clean up a bit, but wanted to get this out there
for comment on more general issues, etc...

To test, I added a localOnly: true to the cow in the tutorial.
This commit is contained in:
David Kelly 2016-07-01 19:27:00 -07:00
parent ae54399177
commit 600348bf10
3 changed files with 262 additions and 26 deletions

View file

@ -14,6 +14,8 @@
#include <sys/stat.h>
#include <glm/glm.hpp>
#include <glm/gtx/norm.hpp>
#include <glm/gtx/vector_angle.hpp>
#ifdef __APPLE__
#include <CoreAudio/AudioHardware.h>
@ -42,6 +44,8 @@
#include <Transform.h>
#include "AudioInjector.h"
#include "AudioLimiter.h"
#include "AudioHRTF.h"
#include "AudioConstants.h"
#include "PositionalAudioStream.h"
#include "AudioClientLogging.h"
@ -852,33 +856,43 @@ void AudioClient::setIsStereoInput(bool isStereoInput) {
bool AudioClient::outputLocalInjector(bool isStereo, AudioInjector* injector) {
if (injector->getLocalBuffer()) {
QAudioFormat localFormat = _desiredOutputFormat;
localFormat.setChannelCount(isStereo ? 2 : 1);
if (injector->getLocalBuffer() && _audioInput ) {
if(isStereo) {
QAudioFormat localFormat = _desiredOutputFormat;
localFormat.setChannelCount(isStereo ? 2 : 1);
QAudioOutput* localOutput = new QAudioOutput(getNamedAudioDeviceForMode(QAudio::AudioOutput, _outputAudioDeviceName),
localFormat,
injector->getLocalBuffer());
QAudioOutput* localOutput = new QAudioOutput(getNamedAudioDeviceForMode(QAudio::AudioOutput, _outputAudioDeviceName),
localFormat,
injector->getLocalBuffer());
// move the localOutput to the same thread as the local injector buffer
localOutput->moveToThread(injector->getLocalBuffer()->thread());
// move the localOutput to the same thread as the local injector buffer
localOutput->moveToThread(injector->getLocalBuffer()->thread());
// have it be stopped when that local buffer is about to close
// We don't want to stop this localOutput and injector whenever this AudioClient singleton goes idle,
// only when the localOutput does. But the connection is to localOutput, so that it happens on the right thread.
connect(localOutput, &QAudioOutput::stateChanged, localOutput, [=](QAudio::State state) {
if (state == QAudio::IdleState) {
localOutput->stop();
injector->stop();
}
});
// have it be stopped when that local buffer is about to close
// We don't want to stop this localOutput and injector whenever this AudioClient singleton goes idle,
// only when the localOutput does. But the connection is to localOutput, so that it happens on the right thread.
connect(localOutput, &QAudioOutput::stateChanged, localOutput, [=](QAudio::State state) {
if (state == QAudio::IdleState) {
localOutput->stop();
injector->stop();
}
});
connect(injector->getLocalBuffer(), &QIODevice::aboutToClose, localOutput, &QAudioOutput::stop);
connect(injector->getLocalBuffer(), &QIODevice::aboutToClose, localOutput, &QAudioOutput::stop);
qCDebug(audioclient) << "Starting QAudioOutput for local injector" << localOutput;
qCDebug(audioclient) << "Starting QAudioOutput for local injector" << localOutput;
localOutput->start(injector->getLocalBuffer());
return localOutput->state() == QAudio::ActiveState;
} else {
// just add it to the vector of active local injectors
// TODO: deal with concurrency perhaps? Maybe not
qDebug() << "adding new injector!!!!!!!";
_activeLocalAudioInjectors.append(injector);
return true;
}
localOutput->start(injector->getLocalBuffer());
return localOutput->state() == QAudio::ActiveState;
}
return false;
@ -1134,20 +1148,225 @@ float AudioClient::getAudioOutputMsecsUnplayed() const {
return msecsAudioOutputUnplayed;
}
void AudioClient::AudioOutputIODevice::renderHRTF(AudioHRTF& hrtf, int16_t* data, float* hrtfBuffer, float azimuth, float gain, qint64 numSamples) {
qint64 numFrames = numSamples/AudioConstants::NETWORK_FRAME_SAMPLES_PER_CHANNEL;
int16_t* dataPtr = data;
float* hrtfPtr = hrtfBuffer;
for(qint64 i=0; i < numFrames; i++) {
hrtf.render(dataPtr, hrtfPtr, 1, azimuth, gain, AudioConstants::NETWORK_FRAME_SAMPLES_PER_CHANNEL);
//qDebug() << "processed frame " << i;
dataPtr += AudioConstants::NETWORK_FRAME_SAMPLES_PER_CHANNEL;
hrtfPtr += 2*AudioConstants::NETWORK_FRAME_SAMPLES_PER_CHANNEL;
}
assert(dataPtr - data <= numSamples);
assert(hrtfPtr - hrtfBuffer <= 2*numSamples);
}
float AudioClient::azimuthForSource(const glm::vec3& relativePosition) {
// copied from AudioMixer, more or less
glm::quat inverseOrientation = glm::inverse(_orientationGetter());
// compute sample delay for the 2 ears to create phase panning
glm::vec3 rotatedSourcePosition = inverseOrientation * relativePosition;
// project the rotated source position vector onto x-y plane
rotatedSourcePosition.y = 0.0f;
static const float SOURCE_DISTANCE_THRESHOLD = 1e-30f;
if (glm::length2(rotatedSourcePosition) > SOURCE_DISTANCE_THRESHOLD) {
// produce an oriented angle about the y-axis
return glm::orientedAngle(glm::vec3(0.0f, 0.0f, -1.0f), glm::normalize(rotatedSourcePosition), glm::vec3(0.0f, -1.0f, 0.0f));
} else {
// no azimuth if they are in same spot
return 0.0f;
}
}
float AudioClient::gainForSource(const glm::vec3& relativePosition, float volume) {
// TODO: put these in a place where we can share with AudioMixer!
const float LOUDNESS_TO_DISTANCE_RATIO = 0.00001f;
const float DEFAULT_ATTENUATION_PER_DOUBLING_IN_DISTANCE = 0.18f;
const float DEFAULT_NOISE_MUTING_THRESHOLD = 0.003f;
const float ATTENUATION_BEGINS_AT_DISTANCE = 1.0f;
//qDebug() << "initial gain is " << volume;
// I'm assuming that the AudioMixer's getting of the stream's attenuation
// factor is basically same as getting volume
float gain = volume;
float distanceBetween = glm::length(relativePosition);
if (distanceBetween < EPSILON ) {
distanceBetween = EPSILON;
}
// audio mixer has notion of zones. Unsure how to map that across here...
// attenuate based on distance now
if (distanceBetween >= ATTENUATION_BEGINS_AT_DISTANCE) {
float distanceCoefficient = 1.0f - (logf(distanceBetween/ATTENUATION_BEGINS_AT_DISTANCE) / logf(2.0f)
* DEFAULT_ATTENUATION_PER_DOUBLING_IN_DISTANCE);
if (distanceCoefficient < 0.0f) {
distanceCoefficient = 0.0f;
}
gain *= distanceCoefficient;
}
//qDebug() << "calculated gain as " << gain;
return gain;
}
qint64 AudioClient::AudioOutputIODevice::readData(char * data, qint64 maxSize) {
auto samplesRequested = maxSize / sizeof(int16_t);
int samplesPopped;
int bytesWritten;
// limit the number of NETWORK_FRAME_SAMPLES_PER_CHANNEL to return regardless
// of what maxSize requests.
static const qint64 MAX_FRAMES = 256;
static float hrtfBuffer[AudioConstants::NETWORK_FRAME_SAMPLES_PER_CHANNEL*MAX_FRAMES];
static int16_t scratchBuffer[AudioConstants::NETWORK_FRAME_SAMPLES_PER_CHANNEL*MAX_FRAMES];
// injectors are 24Khz, use this to limit the injector audio only
static AudioLimiter audioLimiter(AudioConstants::SAMPLE_RATE, AudioConstants::STEREO);
// final output is 48Khz, so use this to limit network audio plus upsampled injectors
static AudioLimiter finalAudioLimiter(2*AudioConstants::SAMPLE_RATE, AudioConstants::STEREO);
// Injectors are 24khz, but we are running at 48Khz here, so need an AudioSRC to upsample
static AudioSRC audioSRC(AudioConstants::SAMPLE_RATE, 2*AudioConstants::SAMPLE_RATE, AudioConstants::STEREO);
// limit maxSize
if (maxSize > AudioConstants::NETWORK_FRAME_SAMPLES_PER_CHANNEL * MAX_FRAMES) {
maxSize = AudioConstants::NETWORK_FRAME_SAMPLES_PER_CHANNEL * MAX_FRAMES;
}
// initialize stuff
memset(hrtfBuffer, 0, sizeof(hrtfBuffer));
QVector<AudioInjector*> injectorsToRemove;
qint64 framesReadFromInjectors = 0;
//qDebug() << "maxSize=" << maxSize << "data ptr: "<< (qint64)data;
for (AudioInjector* injector : _audio->getActiveLocalAudioInjectors()) {
// pop off maxSize/4 (since it is mono, and 24Khz instead of 48Khz)
// bytes from the injector's localBuffer, using the data ptr as a temporary buffer
// note we are only hrtf-ing mono buffers, so we request half of what maxSize is, since it
// wants that many stereo bytes
if (injector->getLocalBuffer() ) {
glm::vec3 relativePosition = injector->getPosition() - _audio->_positionGetter();
qint64 bytesReadFromInjector = injector->getLocalBuffer()->readData(data, maxSize/4);
qint64 framesReadFromInjector = bytesReadFromInjector/sizeof(int16_t);
if (framesReadFromInjector > framesReadFromInjectors) {
framesReadFromInjectors = framesReadFromInjector;
}
if (framesReadFromInjector > 0) {
//qDebug() << "AudioClient::AudioOutputIODevice::readData found " << framesReadFromInjector << " frames";
// calculate these shortly
float gain = _audio->gainForSource(relativePosition, injector->getVolume());
float azimuth = _audio->azimuthForSource(relativePosition);
// now hrtf this guy, one NETWORK_FRAME_SAMPLES_PER_CHANNEL at a time
this->renderHRTF(injector->getLocalHRTF(), (int16_t*)data, hrtfBuffer, azimuth, gain, framesReadFromInjector);
} else {
// probably need to be more clever here than just getting rid of the injector?
//qDebug() << "AudioClient::AudioOutputIODevice::readData no more data, adding to list of injectors to remove";
injectorsToRemove.append(injector);
injector->finish();
}
} else {
// probably need to be more clever here then just getting rid of injector?
//qDebug() << "AudioClient::AudioOutputIODevice::readData injector " << injector << " has no local buffer, adding to list of injectors to remove";
injectorsToRemove.append(injector);
injector->finish();
}
}
if (framesReadFromInjectors > 0) {
// resample the 24Khz injector audio to 48Khz
// but first, hit with limiter :(
audioLimiter.render(hrtfBuffer, (int16_t*)data, framesReadFromInjectors);
audioSRC.render((int16_t*)data, scratchBuffer, framesReadFromInjectors);
// now, lets move this back into the hrtfBuffer
for(int i=0; i<framesReadFromInjectors*4; i++) {
hrtfBuffer[i] = (float)(scratchBuffer[i]) * 1/32676.0f;
}
//qDebug() << "upsampled " << framesReadFromInjectors*2 << " frames, stereo";
}
// k get rid of finished injectors
for (AudioInjector* injector : injectorsToRemove) {
// TODO: I'd expect concurrency issues here -- fix?
_audio->getActiveLocalAudioInjectors().removeOne(injector);
//qDebug() << "removed injector " << injector << " from active injector list!";
}
// now grab the stuff from the network, and mix it in...
if ((samplesPopped = _receivedAudioStream.popSamples((int)samplesRequested, false)) > 0) {
AudioRingBuffer::ConstIterator lastPopOutput = _receivedAudioStream.getLastPopOutput();
lastPopOutput.readSamples((int16_t*)data, samplesPopped);
//qDebug() << "AudioClient::AudioOutputIODevice::readData popped " << samplesPopped << "samples" << "to " << (qint64)data;
bytesWritten = samplesPopped * sizeof(int16_t);
} else {
if (framesReadFromInjectors > 0) {
int16_t* dataPtr = (int16_t*)data;
// convert audio to floats for mixing with limiter...
//qDebug() << "AudioClient::AudioOUtputIODevice::readData converting to floats...";
for (int i=0; i<samplesPopped; i++) {
hrtfBuffer[i] += (float)(*dataPtr++) * 1/32676.0f;
}
// now use limiter, outputting to the data buffer.
// TODO:Since there
// are possibilities for pulling less data from the network than
// we got from injectors (or more), we need to look at being
// more clever here. If we get more bytes here than in the injectors, I'd assume
// that is ok -- we are at the 'end' of the injector sound(s). But
// if we get less here than at the injectors, we maybe want to ponder
// what to do.
//qDebug() << "calling limiter...";
finalAudioLimiter.render(hrtfBuffer, (int16_t*)data, samplesPopped/2);
// the 2 is because the htrf turned the mono into stereo, to match
// what the network data has (2 channels)
if (8*framesReadFromInjectors > bytesWritten) {
bytesWritten = 8*framesReadFromInjectors;
}
//qDebug() << "done, bytes written = " << bytesWritten;
}
} else if (framesReadFromInjectors == 0) {
// if nobody has data, 0 out buffer...
memset(data, 0, maxSize);
bytesWritten = maxSize;
} else {
// only sound from injectors, multiply by 2 as the htrf is stereo
finalAudioLimiter.render(hrtfBuffer, (int16_t*)data, framesReadFromInjectors*2);
bytesWritten = 8*framesReadFromInjectors;
}
int bytesAudioOutputUnplayed = _audio->_audioOutput->bufferSize() - _audio->_audioOutput->bytesFree();
if (!bytesAudioOutputUnplayed) {
qCDebug(audioclient) << "empty audio buffer";

View file

@ -37,6 +37,7 @@
#include <SettingHandle.h>
#include <Sound.h>
#include <StDev.h>
#include <AudioHRTF.h>
#include "AudioIOStats.h"
#include "AudioNoiseGate.h"
@ -86,12 +87,12 @@ public:
void stop() { close(); }
qint64 readData(char * data, qint64 maxSize);
qint64 writeData(const char * data, qint64 maxSize) { return 0; }
int getRecentUnfulfilledReads() { int unfulfilledReads = _unfulfilledReads; _unfulfilledReads = 0; return unfulfilledReads; }
private:
MixedProcessedAudioStream& _receivedAudioStream;
AudioClient* _audio;
int _unfulfilledReads;
void renderHRTF(AudioHRTF& hrtf, int16_t* data, float* hrtfBuffer, float azimuth, float gain, qint64 numSamples);
};
const MixedProcessedAudioStream& getReceivedAudioStream() const { return _receivedAudioStream; }
@ -124,6 +125,8 @@ public:
void setPositionGetter(AudioPositionGetter positionGetter) { _positionGetter = positionGetter; }
void setOrientationGetter(AudioOrientationGetter orientationGetter) { _orientationGetter = orientationGetter; }
QVector<AudioInjector*>& getActiveLocalAudioInjectors() { return _activeLocalAudioInjectors; }
static const float CALLBACK_ACCELERATOR_RATIO;
@ -291,6 +294,12 @@ private:
void checkDevices();
bool _hasReceivedFirstPacket = false;
QVector<AudioInjector*> _activeLocalAudioInjectors;
float azimuthForSource(const glm::vec3& relativePosition);
float gainForSource(const glm::vec3& relativePosition, float volume);
};

View file

@ -26,6 +26,7 @@
#include "AudioInjectorLocalBuffer.h"
#include "AudioInjectorOptions.h"
#include "AudioHRTF.h"
#include "Sound.h"
class AbstractAudioInterface;
@ -54,8 +55,11 @@ public:
void setCurrentSendOffset(int currentSendOffset) { _currentSendOffset = currentSendOffset; }
AudioInjectorLocalBuffer* getLocalBuffer() const { return _localBuffer; }
AudioHRTF& getLocalHRTF() { return _localHRTF; }
bool isLocalOnly() const { return _options.localOnly; }
float getVolume() const { return _options.volume; }
glm::vec3 getPosition() const { return _options.position; }
void setLocalAudioInterface(AbstractAudioInterface* localAudioInterface) { _localAudioInterface = localAudioInterface; }
static AudioInjector* playSoundAndDelete(const QByteArray& buffer, const AudioInjectorOptions options, AbstractAudioInterface* localInterface);
@ -74,14 +78,15 @@ public slots:
float getLoudness() const { return _loudness; }
bool isPlaying() const { return _state == State::NotFinished || _state == State::NotFinishedWithPendingDelete; }
void finish();
signals:
void finished();
void restarting();
private slots:
/*private slots:
void finish();
*/
private:
void setupInjection();
int64_t injectNextFrame();
@ -103,6 +108,9 @@ private:
std::unique_ptr<QElapsedTimer> _frameTimer { nullptr };
quint16 _outgoingSequenceNumber { 0 };
// when the injector is local, we need this
AudioHRTF _localHRTF;
// for local injection,
friend class AudioInjectorManager;
};