862 lines · 36.2 KB · C++
RawDownload
////  Audio.cpp//  interface////  Created by Stephen Birarda on 1/22/13.//  Copyright (c) 2013 High Fidelity, Inc. All rights reserved.// #include <cstring>#include <sys/stat.h> #include <math.h> #ifdef __APPLE__#include <CoreAudio/AudioHardware.h>#endif #include <QtCore/QBuffer>#include <QtMultimedia/QAudioInput>#include <QtMultimedia/QAudioOutput>#include <QSvgRenderer> #include <AngleUtil.h>#include <NodeList.h>#include <PacketHeaders.h>#include <SharedUtil.h>#include <StdDev.h>#include <UUID.h> #include "Application.h"#include "Audio.h"#include "Menu.h"#include "Util.h" static const float JITTER_BUFFER_LENGTH_MSECS = 12;static const short JITTER_BUFFER_SAMPLES = JITTER_BUFFER_LENGTH_MSECS * NUM_AUDIO_CHANNELS * (SAMPLE_RATE / 1000.0); static const float AUDIO_CALLBACK_MSECS = (float) NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL / (float)SAMPLE_RATE * 1000.0; static const int NUMBER_OF_NOISE_SAMPLE_FRAMES = 300; // Mute icon configrationstatic const int ICON_SIZE = 24;static const int ICON_LEFT = 0;static const int ICON_TOP = 115;static const int ICON_TOP_MIRROR = 220; Audio::Audio(Oscilloscope* scope, int16_t initialJitterBufferSamples, QObject* parent) :    AbstractAudioInterface(parent),    _audioInput(NULL),    _desiredInputFormat(),    _inputFormat(),    _numInputCallbackBytes(0),    _audioOutput(NULL),    _desiredOutputFormat(),    _outputFormat(),    _outputDevice(NULL),    _numOutputCallbackBytes(0),    _loopbackAudioOutput(NULL),    _loopbackOutputDevice(NULL),    _proceduralAudioOutput(NULL),    _proceduralOutputDevice(NULL),    _inputRingBuffer(0),    _ringBuffer(NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL),    _scope(scope),    _averagedLatency(0.0),    _measuredJitter(0),    _jitterBufferSamples(initialJitterBufferSamples),    _lastInputLoudness(0),    _dcOffset(0),    _noiseGateMeasuredFloor(0),    _noiseGateSampleCounter(0),    _noiseGateOpen(false),    _noiseGateEnabled(true),    _noiseGateFramesToClose(0),    _lastVelocity(0),    _lastAcceleration(0),    _totalPacketsReceived(0),    _collisionSoundMagnitude(0.0f),    _collisionSoundFrequency(0.0f),    _collisionSoundNoise(0.0f),    _collisionSoundDuration(0.0f),    _proceduralEffectSample(0),    _numFramesDisplayStarve(0),    _muted(false){    // clear the array of locally injected samples    memset(_localProceduralSamples, 0, NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL);    // Create the noise sample array    _noiseSampleFrames = new float[NUMBER_OF_NOISE_SAMPLE_FRAMES];} void Audio::init(QGLWidget *parent) {    switchToResourcesParentIfRequired();    _micTextureId = parent->bindTexture(QImage("./resources/images/mic.svg"));    _muteTextureId = parent->bindTexture(QImage("./resources/images/mute.svg"));} void Audio::reset() {    _ringBuffer.reset();} QAudioDeviceInfo defaultAudioDeviceForMode(QAudio::Mode mode) {#ifdef __APPLE__    if (QAudioDeviceInfo::availableDevices(mode).size() > 1) {        AudioDeviceID defaultDeviceID = 0;        uint32_t propertySize = sizeof(AudioDeviceID);        AudioObjectPropertyAddress propertyAddress = {            kAudioHardwarePropertyDefaultInputDevice,            kAudioObjectPropertyScopeGlobal,            kAudioObjectPropertyElementMaster        };         if (mode == QAudio::AudioOutput) {            propertyAddress.mSelector = kAudioHardwarePropertyDefaultOutputDevice;        }          OSStatus getPropertyError = AudioObjectGetPropertyData(kAudioObjectSystemObject,                                                               &propertyAddress,                                                               0,                                                               NULL,                                                               &propertySize,                                                               &defaultDeviceID);         if (!getPropertyError && propertySize) {            CFStringRef deviceName = NULL;            propertySize = sizeof(deviceName);            propertyAddress.mSelector = kAudioDevicePropertyDeviceNameCFString;            getPropertyError = AudioObjectGetPropertyData(defaultDeviceID, &propertyAddress, 0,                                                          NULL, &propertySize, &deviceName);             if (!getPropertyError && propertySize) {                // find a device in the list that matches the name we have and return it                foreach(QAudioDeviceInfo audioDevice, QAudioDeviceInfo::availableDevices(mode)) {                    if (audioDevice.deviceName() == CFStringGetCStringPtr(deviceName, kCFStringEncodingMacRoman)) {                        return audioDevice;                    }                }            }        }    }#endif     // fallback for failed lookup is the default device    return (mode == QAudio::AudioInput) ? QAudioDeviceInfo::defaultInputDevice() : QAudioDeviceInfo::defaultOutputDevice();} bool adjustedFormatForAudioDevice(const QAudioDeviceInfo& audioDevice,                                  const QAudioFormat& desiredAudioFormat,                                  QAudioFormat& adjustedAudioFormat) {    if (!audioDevice.isFormatSupported(desiredAudioFormat)) {        qDebug() << "The desired format for audio I/O is" << desiredAudioFormat;        qDebug("The desired audio format is not supported by this device");                if (desiredAudioFormat.channelCount() == 1) {            adjustedAudioFormat = desiredAudioFormat;            adjustedAudioFormat.setChannelCount(2);             if (audioDevice.isFormatSupported(adjustedAudioFormat)) {                return true;            } else {                adjustedAudioFormat.setChannelCount(1);            }        }         if (audioDevice.supportedSampleRates().contains(SAMPLE_RATE * 2)) {            // use 48, which is a sample downsample, upsample            adjustedAudioFormat = desiredAudioFormat;            adjustedAudioFormat.setSampleRate(SAMPLE_RATE * 2);             // return the nearest in case it needs 2 channels            adjustedAudioFormat = audioDevice.nearestFormat(adjustedAudioFormat);            return true;        }         return false;    } else {        // set the adjustedAudioFormat to the desiredAudioFormat, since it will work        adjustedAudioFormat = desiredAudioFormat;        return true;    }} void linearResampling(int16_t* sourceSamples, int16_t* destinationSamples,                      unsigned int numSourceSamples, unsigned int numDestinationSamples,                      const QAudioFormat& sourceAudioFormat, const QAudioFormat& destinationAudioFormat) {    if (sourceAudioFormat == destinationAudioFormat) {        memcpy(destinationSamples, sourceSamples, numSourceSamples * sizeof(int16_t));    } else {        float sourceToDestinationFactor = (sourceAudioFormat.sampleRate() / (float) destinationAudioFormat.sampleRate())            * (sourceAudioFormat.channelCount() / (float) destinationAudioFormat.channelCount());         // take into account the number of channels in source and destination        // accomodate for the case where have an output with > 2 channels        // this is the case with our HDMI capture         if (sourceToDestinationFactor >= 2) {            // we need to downsample from 48 to 24            // for now this only supports a mono output - this would be the case for audio input             for (unsigned int i = sourceAudioFormat.channelCount(); i < numSourceSamples; i += 2 * sourceAudioFormat.channelCount()) {                if (i + (sourceAudioFormat.channelCount()) >= numSourceSamples) {                    destinationSamples[(i - sourceAudioFormat.channelCount()) / (int) sourceToDestinationFactor] =                        (sourceSamples[i - sourceAudioFormat.channelCount()] / 2)                        + (sourceSamples[i] / 2);                } else {                    destinationSamples[(i - sourceAudioFormat.channelCount()) / (int) sourceToDestinationFactor] =                        (sourceSamples[i - sourceAudioFormat.channelCount()] / 4)                        + (sourceSamples[i] / 2)                        + (sourceSamples[i + sourceAudioFormat.channelCount()] / 4);                }            }         } else {            // upsample from 24 to 48            // for now this only supports a stereo to stereo conversion - this is our case for network audio to output            int sourceIndex = 0;            int dtsSampleRateFactor = (destinationAudioFormat.sampleRate() / sourceAudioFormat.sampleRate());            int sampleShift = destinationAudioFormat.channelCount() * dtsSampleRateFactor;            int destinationToSourceFactor = (1 / sourceToDestinationFactor);             for (unsigned int i = 0; i < numDestinationSamples; i += sampleShift) {                sourceIndex = (i / destinationToSourceFactor);                 // fill the L/R channels and make the rest silent                for (unsigned int j = i; j < i + sampleShift; j++) {                    if (j % destinationAudioFormat.channelCount() == 0) {                        // left channel                        destinationSamples[j] = sourceSamples[sourceIndex];                    } else if (j % destinationAudioFormat.channelCount() == 1) {                         // right channel                        destinationSamples[j] = sourceSamples[sourceIndex + (sourceAudioFormat.channelCount() > 1 ? 1 : 0)];                    } else {                        // channels above 2, fill with silence                        destinationSamples[j] = 0;                    }                }            }        }    }} const int CALLBACK_ACCELERATOR_RATIO = 2; void Audio::start() {     // set up the desired audio format    _desiredInputFormat.setSampleRate(SAMPLE_RATE);    _desiredInputFormat.setSampleSize(16);    _desiredInputFormat.setCodec("audio/pcm");    _desiredInputFormat.setSampleType(QAudioFormat::SignedInt);    _desiredInputFormat.setByteOrder(QAudioFormat::LittleEndian);    _desiredInputFormat.setChannelCount(1);     _desiredOutputFormat = _desiredInputFormat;    _desiredOutputFormat.setChannelCount(2);     QAudioDeviceInfo inputDeviceInfo = defaultAudioDeviceForMode(QAudio::AudioInput);    qDebug() << "The audio input device is" << inputDeviceInfo.deviceName();        if (adjustedFormatForAudioDevice(inputDeviceInfo, _desiredInputFormat, _inputFormat)) {        qDebug() << "The format to be used for audio input is" << _inputFormat;                _audioInput = new QAudioInput(inputDeviceInfo, _inputFormat, this);        _numInputCallbackBytes = NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL * _inputFormat.channelCount()            * (_inputFormat.sampleRate() / SAMPLE_RATE)            / CALLBACK_ACCELERATOR_RATIO;        _audioInput->setBufferSize(_numInputCallbackBytes);         QAudioDeviceInfo outputDeviceInfo = defaultAudioDeviceForMode(QAudio::AudioOutput);        qDebug() << "The audio output device is" << outputDeviceInfo.deviceName();         if (adjustedFormatForAudioDevice(outputDeviceInfo, _desiredOutputFormat, _outputFormat)) {            qDebug() << "The format to be used for audio output is" << _outputFormat;                        _inputRingBuffer.resizeForFrameSize(_numInputCallbackBytes * CALLBACK_ACCELERATOR_RATIO / sizeof(int16_t));            _inputDevice = _audioInput->start();            connect(_inputDevice, SIGNAL(readyRead()), this, SLOT(handleAudioInput()));             // setup our general output device for audio-mixer audio            _audioOutput = new QAudioOutput(outputDeviceInfo, _outputFormat, this);            _outputDevice = _audioOutput->start();             // setup a loopback audio output device            _loopbackAudioOutput = new QAudioOutput(outputDeviceInfo, _outputFormat, this);                        // setup a procedural audio output device            _proceduralAudioOutput = new QAudioOutput(outputDeviceInfo, _outputFormat, this);             gettimeofday(&_lastReceiveTime, NULL);        }         return;    }        qDebug() << "Unable to set up audio I/O because of a problem with input or output formats.";} void Audio::handleAudioInput() {    static char monoAudioDataPacket[MAX_PACKET_SIZE];     static int numBytesPacketHeader = numBytesForPacketHeaderGivenPacketType(PacketTypeMicrophoneAudioNoEcho);    static int leadingBytes = numBytesPacketHeader + sizeof(glm::vec3) + sizeof(glm::quat);     static int16_t* monoAudioSamples = (int16_t*) (monoAudioDataPacket + leadingBytes);     static float inputToNetworkInputRatio = _numInputCallbackBytes * CALLBACK_ACCELERATOR_RATIO        / NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL;     static unsigned int inputSamplesRequired = NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL * inputToNetworkInputRatio;     QByteArray inputByteArray = _inputDevice->readAll();     if (Menu::getInstance()->isOptionChecked(MenuOption::EchoLocalAudio) && !_muted) {        // if this person wants local loopback add that to the locally injected audio         if (!_loopbackOutputDevice) {            // we didn't have the loopback output device going so set that up now            _loopbackOutputDevice = _loopbackAudioOutput->start();        }         if (_inputFormat == _outputFormat) {            _loopbackOutputDevice->write(inputByteArray);        } else {            static float loopbackOutputToInputRatio = (_outputFormat.sampleRate() / (float) _inputFormat.sampleRate())                * (_outputFormat.channelCount() / _inputFormat.channelCount());             QByteArray loopBackByteArray(inputByteArray.size() * loopbackOutputToInputRatio, 0);             linearResampling((int16_t*) inputByteArray.data(), (int16_t*) loopBackByteArray.data(),                             inputByteArray.size() / sizeof(int16_t),                             loopBackByteArray.size() / sizeof(int16_t), _inputFormat, _outputFormat);             _loopbackOutputDevice->write(loopBackByteArray);        }    }     _inputRingBuffer.writeData(inputByteArray.data(), inputByteArray.size());     while (_inputRingBuffer.samplesAvailable() > inputSamplesRequired) {         int16_t* inputAudioSamples = new int16_t[inputSamplesRequired];        _inputRingBuffer.readSamples(inputAudioSamples, inputSamplesRequired);         // zero out the monoAudioSamples array and the locally injected audio        memset(monoAudioSamples, 0, NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL);         // zero out the locally injected audio in preparation for audio procedural sounds        memset(_localProceduralSamples, 0, NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL);         if (!_muted) {            // we aren't muted, downsample the input audio            linearResampling((int16_t*) inputAudioSamples,                             monoAudioSamples,                             inputSamplesRequired,                             NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL,                             _inputFormat, _desiredInputFormat);             //            //  Impose Noise Gate            //            //  The Noise Gate is used to reject constant background noise by measuring the noise            //  floor observed at the microphone and then opening the 'gate' to allow microphone            //  signals to be transmitted when the microphone samples average level exceeds a multiple            //  of the noise floor.            //            //  NOISE_GATE_HEIGHT:  How loud you have to speak relative to noise background to open the gate.            //                      Make this value lower for more sensitivity and less rejection of noise.            //  NOISE_GATE_WIDTH:   The number of samples in an audio frame for which the height must be exceeded            //                      to open the gate.            //  NOISE_GATE_CLOSE_FRAME_DELAY:  Once the noise is below the gate height for the frame, how many frames            //                      will we wait before closing the gate.            //  NOISE_GATE_FRAMES_TO_AVERAGE:  How many audio frames should we average together to compute noise floor.            //                      More means better rejection but also can reject continuous things like singing.            // NUMBER_OF_NOISE_SAMPLE_FRAMES:  How often should we re-evaluate the noise floor?                         float loudness = 0;            float thisSample = 0;            int samplesOverNoiseGate = 0;                        const float NOISE_GATE_HEIGHT = 7.f;            const int NOISE_GATE_WIDTH = 5;            const int NOISE_GATE_CLOSE_FRAME_DELAY = 5;            const int NOISE_GATE_FRAMES_TO_AVERAGE = 5;            const float DC_OFFSET_AVERAGING = 0.99f;                        float measuredDcOffset = 0.f;                        for (int i = 0; i < NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL; i++) {                measuredDcOffset += monoAudioSamples[i];                monoAudioSamples[i] -= (int16_t) _dcOffset;                thisSample = fabsf(monoAudioSamples[i]);                loudness += thisSample;                //  Noise Reduction:  Count peaks above the average loudness                if (thisSample > (_noiseGateMeasuredFloor * NOISE_GATE_HEIGHT)) {                    samplesOverNoiseGate++;                }            }                        measuredDcOffset /= NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL;            if (_dcOffset == 0.f) {                // On first frame, copy over measured offset                _dcOffset = measuredDcOffset;            } else {                _dcOffset = DC_OFFSET_AVERAGING * _dcOffset + (1.f - DC_OFFSET_AVERAGING) * measuredDcOffset;            }                        //            _lastInputLoudness = fabs(loudness / NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL);                        float averageOfAllSampleFrames = 0.f;            _noiseSampleFrames[_noiseGateSampleCounter++] = _lastInputLoudness;            if (_noiseGateSampleCounter == NUMBER_OF_NOISE_SAMPLE_FRAMES) {                float smallestSample = FLT_MAX;                for (int i = 0; i <= NUMBER_OF_NOISE_SAMPLE_FRAMES - NOISE_GATE_FRAMES_TO_AVERAGE; i+= NOISE_GATE_FRAMES_TO_AVERAGE) {                    float thisAverage = 0.0f;                    for (int j = i; j < i + NOISE_GATE_FRAMES_TO_AVERAGE; j++) {                        thisAverage += _noiseSampleFrames[j];                        averageOfAllSampleFrames += _noiseSampleFrames[j];                    }                    thisAverage /= NOISE_GATE_FRAMES_TO_AVERAGE;                                        if (thisAverage < smallestSample) {                        smallestSample = thisAverage;                    }                }                averageOfAllSampleFrames /= NUMBER_OF_NOISE_SAMPLE_FRAMES;                _noiseGateMeasuredFloor = smallestSample;                _noiseGateSampleCounter = 0;             }             if (_noiseGateEnabled) {                if (samplesOverNoiseGate > NOISE_GATE_WIDTH) {                    _noiseGateOpen = true;                    _noiseGateFramesToClose = NOISE_GATE_CLOSE_FRAME_DELAY;                } else {                    if (--_noiseGateFramesToClose == 0) {                        _noiseGateOpen = false;                    }                }                if (!_noiseGateOpen) {                    for (int i = 0; i < NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL; i++) {                        monoAudioSamples[i] = 0;                    }                }            }             // add input data just written to the scope            QMetaObject::invokeMethod(_scope, "addSamples", Qt::QueuedConnection,                                      Q_ARG(QByteArray, QByteArray((char*) monoAudioSamples,                                                                   NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL)),                                      Q_ARG(bool, false), Q_ARG(bool, true));        } else {            // our input loudness is 0, since we're muted            _lastInputLoudness = 0;        }         // add procedural effects to the appropriate input samples        addProceduralSounds(monoAudioSamples,                            NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL);                if (!_proceduralOutputDevice) {            _proceduralOutputDevice = _proceduralAudioOutput->start();        }                // send whatever procedural sounds we want to locally loop back to the _proceduralOutputDevice        QByteArray proceduralOutput;        proceduralOutput.resize(NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL * 4 * sizeof(int16_t));                linearResampling(_localProceduralSamples,                         reinterpret_cast<int16_t*>(proceduralOutput.data()),                         NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL,                         NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL * 4,                         _desiredInputFormat, _outputFormat);                _proceduralOutputDevice->write(proceduralOutput);         NodeList* nodeList = NodeList::getInstance();        SharedNodePointer audioMixer = nodeList->soloNodeOfType(NodeType::AudioMixer);                if (audioMixer && audioMixer->getActiveSocket()) {            MyAvatar* interfaceAvatar = Application::getInstance()->getAvatar();            glm::vec3 headPosition = interfaceAvatar->getHead()->getPosition();            glm::quat headOrientation = interfaceAvatar->getHead()->getOrientation();             // we need the amount of bytes in the buffer + 1 for type            // + 12 for 3 floats for position + float for bearing + 1 attenuation byte             PacketType packetType = Menu::getInstance()->isOptionChecked(MenuOption::EchoServerAudio)                ? PacketTypeMicrophoneAudioWithEcho : PacketTypeMicrophoneAudioNoEcho;             char* currentPacketPtr = monoAudioDataPacket + populatePacketHeader(monoAudioDataPacket, packetType);             // memcpy the three float positions            memcpy(currentPacketPtr, &headPosition, sizeof(headPosition));            currentPacketPtr += (sizeof(headPosition)); 

Lines 1–500 of 862.

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t