void AudioRecorder::processBuffer(const QAudioBuffer& buffer)
{
if (audioLevels.count() != buffer.format().channelCount()) {
qDeleteAll(audioLevels);
audioLevels.clear();
for (int i = 0; i < buffer.format().channelCount(); ++i) {
QAudioLevel *level = new QAudioLevel(ui->centralwidget);
audioLevels.append(level);
ui->levelsLayout->addWidget(level);
}
}
QVector<qreal> levels = getBufferLevels(buffer);
for (int i = 0; i < levels.count(); ++i)
audioLevels.at(i)->setLevel(levels.at(i));
}
bool WaveFileWriter::write(const QAudioBuffer &buffer)
{
if (buffer.format() != m_format)
return false; // buffer format has changed
qint64 written = file.write((const char *)buffer.constData(), buffer.byteCount());
m_dataLength += written;
return written == buffer.byteCount();
}
void AudioDecoder::bufferReady()
{
// read a buffer from audio decoder
QAudioBuffer buffer = m_decoder.read();
if (!buffer.isValid())
return;
if (!m_fileWriter.isOpen() && !m_fileWriter.open(m_targetFilename, buffer.format())) {
m_decoder.stop();
return;
}
m_fileWriter.write(buffer);
}
void Analyzer::preProcess(const QAudioBuffer &input)
{
m_input.fill(0);
switch (input.format().sampleSize()) {
case 8:
extractAndScale<qint8>(input);
break;
case 16:
extractAndScale<qint16>(input);
break;
case 32:
extractAndScale<qint32>(input);
break;
case 64:
extractAndScale<qint64>(input);
break;
}
// Find a simple least squares fit y = ax + b to the scaled input
const auto xMean = 0.5 * (m_sampleSize + 1);
const auto sum = std::accumulate(m_input.constBegin(), m_input.constBegin() + m_sampleSize, 0);
const auto yMean = sum / m_sampleSize;
qreal covXY = 0; // Cross-covariance
qreal varX = 0; // Variance
auto y = m_input.constBegin();
for (int x = 0; x < m_input.size(); ++x, ++y) {
const auto dx = x - xMean;
covXY += dx * (*y - yMean);
varX += dx * dx;
}
const auto a = covXY / varX;
const auto b = yMean - a * xMean;
// Subtract this fit and apply the window function
auto i = m_input.begin();
auto w = m_window.constBegin();
for (int x = 0; x < m_input.size(); ++i, ++w, ++x)
*i = *w * (*i - (a * x + b));
}
void Analyzer::doAnalysis(const QAudioBuffer &input)
{
if (m_state != Ready)
return;
if (m_calibrateFilter)
setState(CalibratingFilter);
else
setState(Processing);
// Process the bytearray into m_input and store a copy for computation of
// the SNAC function
m_currentFormat = input.format();
preProcess(input);
auto processedInput = m_input;
processedInput.detach();
getSpectrum();
if (m_calibrateFilter)
calibrateFilter();
processSpectrum();
// Finally, compute the normalised ACF and frequency estimate
getAcf();
const auto snac = computeSnac(m_input, processedInput);
const auto snacPeak = determineSnacFundamental(snac);
Spectrum snacPeaks;
if (snacPeak.frequency > 0)
snacPeaks << snacPeak;
// The accuracy of the obtained fundamental is fair, but can be improved
// using the accurate power spectrum stored earlier, which also allows
// identifying overtones
const auto harmonics = findHarmonics(m_spectrum, m_currentFormat.sampleRate() / snacPeak.frequency);
// Report analysis results
setState(Ready);
emit done(harmonics, m_spectrum, snac, snacPeaks);
}
void tst_QAudioDecoderBackend::fileTest()
{
QAudioDecoder d;
QAudioBuffer buffer;
quint64 duration = 0;
int byteCount = 0;
int sampleCount = 0;
QVERIFY(d.state() == QAudioDecoder::StoppedState);
QVERIFY(d.bufferAvailable() == false);
QCOMPARE(d.sourceFilename(), QString(""));
QVERIFY(d.audioFormat() == QAudioFormat());
// Test local file
QFileInfo fileInfo(QFINDTESTDATA(TEST_FILE_NAME));
d.setSourceFilename(fileInfo.absoluteFilePath());
QVERIFY(d.state() == QAudioDecoder::StoppedState);
QVERIFY(!d.bufferAvailable());
QCOMPARE(d.sourceFilename(), fileInfo.absoluteFilePath());
QSignalSpy readySpy(&d, SIGNAL(bufferReady()));
QSignalSpy bufferChangedSpy(&d, SIGNAL(bufferAvailableChanged(bool)));
QSignalSpy errorSpy(&d, SIGNAL(error(QAudioDecoder::Error)));
QSignalSpy stateSpy(&d, SIGNAL(stateChanged(QAudioDecoder::State)));
QSignalSpy durationSpy(&d, SIGNAL(durationChanged(qint64)));
QSignalSpy finishedSpy(&d, SIGNAL(finished()));
QSignalSpy positionSpy(&d, SIGNAL(positionChanged(qint64)));
d.start();
QTRY_VERIFY(d.state() == QAudioDecoder::DecodingState);
QTRY_VERIFY(!stateSpy.isEmpty());
QTRY_VERIFY(!readySpy.isEmpty());
QTRY_VERIFY(!bufferChangedSpy.isEmpty());
QVERIFY(d.bufferAvailable());
QTRY_VERIFY(!durationSpy.isEmpty());
QVERIFY(qAbs(d.duration() - 1000) < 20);
buffer = d.read();
QVERIFY(buffer.isValid());
// Test file is 44.1K 16bit mono, 44094 samples
QCOMPARE(buffer.format().channelCount(), 1);
QCOMPARE(buffer.format().sampleRate(), 44100);
QCOMPARE(buffer.format().sampleSize(), 16);
QCOMPARE(buffer.format().sampleType(), QAudioFormat::SignedInt);
QCOMPARE(buffer.format().codec(), QString("audio/pcm"));
QCOMPARE(buffer.byteCount(), buffer.sampleCount() * 2); // 16bit mono
// The decoder should still have no format set
QVERIFY(d.audioFormat() == QAudioFormat());
QVERIFY(errorSpy.isEmpty());
duration += buffer.duration();
sampleCount += buffer.sampleCount();
byteCount += buffer.byteCount();
// Now drain the decoder
if (sampleCount < 44094) {
QTRY_COMPARE(d.bufferAvailable(), true);
}
while (d.bufferAvailable()) {
buffer = d.read();
QVERIFY(buffer.isValid());
QTRY_VERIFY(!positionSpy.isEmpty());
QVERIFY(positionSpy.takeLast().at(0).toLongLong() == qint64(duration / 1000));
duration += buffer.duration();
sampleCount += buffer.sampleCount();
byteCount += buffer.byteCount();
if (sampleCount < 44094) {
QTRY_COMPARE(d.bufferAvailable(), true);
}
}
// Make sure the duration is roughly correct (+/- 20ms)
QCOMPARE(sampleCount, 44094);
QCOMPARE(byteCount, 44094 * 2);
QVERIFY(qAbs(qint64(duration) - 1000000) < 20000);
QVERIFY(qAbs((d.position() + (buffer.duration() / 1000)) - 1000) < 20);
QTRY_COMPARE(finishedSpy.count(), 1);
QVERIFY(!d.bufferAvailable());
QTRY_COMPARE(d.state(), QAudioDecoder::StoppedState);
d.stop();
QTRY_COMPARE(d.state(), QAudioDecoder::StoppedState);
QTRY_COMPARE(durationSpy.count(), 2);
QCOMPARE(d.duration(), qint64(-1));
QVERIFY(!d.bufferAvailable());
readySpy.clear();
bufferChangedSpy.clear();
stateSpy.clear();
durationSpy.clear();
finishedSpy.clear();
positionSpy.clear();
// change output audio format
QAudioFormat format;
format.setChannelCount(2);
format.setSampleSize(8);
format.setSampleRate(11050);
format.setCodec("audio/pcm");
format.setSampleType(QAudioFormat::SignedInt);
d.setAudioFormat(format);
// We expect 1 second still, at 11050 * 2 samples == 22k samples.
// (at 1 byte/sample -> 22kb)
// Make sure it stuck
QVERIFY(d.audioFormat() == format);
duration = 0;
sampleCount = 0;
byteCount = 0;
d.start();
QTRY_VERIFY(d.state() == QAudioDecoder::DecodingState);
QTRY_VERIFY(!stateSpy.isEmpty());
QTRY_VERIFY(!readySpy.isEmpty());
QTRY_VERIFY(!bufferChangedSpy.isEmpty());
QVERIFY(d.bufferAvailable());
QTRY_VERIFY(!durationSpy.isEmpty());
QVERIFY(qAbs(d.duration() - 1000) < 20);
buffer = d.read();
QVERIFY(buffer.isValid());
// See if we got the right format
QVERIFY(buffer.format() == format);
// The decoder should still have the same format
QVERIFY(d.audioFormat() == format);
QVERIFY(errorSpy.isEmpty());
duration += buffer.duration();
sampleCount += buffer.sampleCount();
byteCount += buffer.byteCount();
// Now drain the decoder
if (duration < 998000) {
QTRY_COMPARE(d.bufferAvailable(), true);
}
while (d.bufferAvailable()) {
buffer = d.read();
QVERIFY(buffer.isValid());
QTRY_VERIFY(!positionSpy.isEmpty());
QVERIFY(positionSpy.takeLast().at(0).toLongLong() == qint64(duration / 1000));
QVERIFY(d.position() - (duration / 1000) < 20);
duration += buffer.duration();
sampleCount += buffer.sampleCount();
byteCount += buffer.byteCount();
if (duration < 998000) {
QTRY_COMPARE(d.bufferAvailable(), true);
}
}
// Resampling might end up with fewer or more samples
// so be a bit sloppy
QVERIFY(qAbs(sampleCount - 22047) < 100);
QVERIFY(qAbs(byteCount - 22047) < 100);
QVERIFY(qAbs(qint64(duration) - 1000000) < 20000);
QVERIFY(qAbs((d.position() + (buffer.duration() / 1000)) - 1000) < 20);
QTRY_COMPARE(finishedSpy.count(), 1);
QVERIFY(!d.bufferAvailable());
QTRY_COMPARE(d.state(), QAudioDecoder::StoppedState);
d.stop();
QTRY_COMPARE(d.state(), QAudioDecoder::StoppedState);
QTRY_COMPARE(durationSpy.count(), 2);
QCOMPARE(d.duration(), qint64(-1));
QVERIFY(!d.bufferAvailable());
}
void tst_QAudioDecoderBackend::deviceTest()
{
QAudioDecoder d;
QAudioBuffer buffer;
quint64 duration = 0;
int sampleCount = 0;
QSignalSpy readySpy(&d, SIGNAL(bufferReady()));
QSignalSpy bufferChangedSpy(&d, SIGNAL(bufferAvailableChanged(bool)));
QSignalSpy errorSpy(&d, SIGNAL(error(QAudioDecoder::Error)));
QSignalSpy stateSpy(&d, SIGNAL(stateChanged(QAudioDecoder::State)));
QSignalSpy durationSpy(&d, SIGNAL(durationChanged(qint64)));
QSignalSpy finishedSpy(&d, SIGNAL(finished()));
QSignalSpy positionSpy(&d, SIGNAL(positionChanged(qint64)));
QVERIFY(d.state() == QAudioDecoder::StoppedState);
QVERIFY(d.bufferAvailable() == false);
QCOMPARE(d.sourceFilename(), QString(""));
QVERIFY(d.audioFormat() == QAudioFormat());
QFileInfo fileInfo(QFINDTESTDATA(TEST_FILE_NAME));
QFile file(fileInfo.absoluteFilePath());
QVERIFY(file.open(QIODevice::ReadOnly));
d.setSourceDevice(&file);
QVERIFY(d.sourceDevice() == &file);
QVERIFY(d.sourceFilename().isEmpty());
// We haven't set the format yet
QVERIFY(d.audioFormat() == QAudioFormat());
d.start();
QTRY_VERIFY(d.state() == QAudioDecoder::DecodingState);
QTRY_VERIFY(!stateSpy.isEmpty());
QTRY_VERIFY(!readySpy.isEmpty());
QTRY_VERIFY(!bufferChangedSpy.isEmpty());
QVERIFY(d.bufferAvailable());
QTRY_VERIFY(!durationSpy.isEmpty());
QVERIFY(qAbs(d.duration() - 1000) < 20);
buffer = d.read();
QVERIFY(buffer.isValid());
// Test file is 44.1K 16bit mono
QCOMPARE(buffer.format().channelCount(), 1);
QCOMPARE(buffer.format().sampleRate(), 44100);
QCOMPARE(buffer.format().sampleSize(), 16);
QCOMPARE(buffer.format().sampleType(), QAudioFormat::SignedInt);
QCOMPARE(buffer.format().codec(), QString("audio/pcm"));
QVERIFY(errorSpy.isEmpty());
duration += buffer.duration();
sampleCount += buffer.sampleCount();
// Now drain the decoder
if (sampleCount < 44094) {
QTRY_COMPARE(d.bufferAvailable(), true);
}
while (d.bufferAvailable()) {
buffer = d.read();
QVERIFY(buffer.isValid());
QTRY_VERIFY(!positionSpy.isEmpty());
QVERIFY(positionSpy.takeLast().at(0).toLongLong() == qint64(duration / 1000));
QVERIFY(d.position() - (duration / 1000) < 20);
duration += buffer.duration();
sampleCount += buffer.sampleCount();
if (sampleCount < 44094) {
QTRY_COMPARE(d.bufferAvailable(), true);
}
}
// Make sure the duration is roughly correct (+/- 20ms)
QCOMPARE(sampleCount, 44094);
QVERIFY(qAbs(qint64(duration) - 1000000) < 20000);
QVERIFY(qAbs((d.position() + (buffer.duration() / 1000)) - 1000) < 20);
QTRY_COMPARE(finishedSpy.count(), 1);
QVERIFY(!d.bufferAvailable());
QTRY_COMPARE(d.state(), QAudioDecoder::StoppedState);
d.stop();
QTRY_COMPARE(d.state(), QAudioDecoder::StoppedState);
QVERIFY(!d.bufferAvailable());
QTRY_COMPARE(durationSpy.count(), 2);
QCOMPARE(d.duration(), qint64(-1));
readySpy.clear();
bufferChangedSpy.clear();
stateSpy.clear();
durationSpy.clear();
finishedSpy.clear();
positionSpy.clear();
// Now try changing formats
QAudioFormat format;
format.setChannelCount(2);
format.setSampleSize(8);
format.setSampleRate(8000);
format.setCodec("audio/pcm");
format.setSampleType(QAudioFormat::SignedInt);
d.setAudioFormat(format);
// Make sure it stuck
QVERIFY(d.audioFormat() == format);
d.start();
QTRY_VERIFY(d.state() == QAudioDecoder::DecodingState);
QTRY_VERIFY(!stateSpy.isEmpty());
QTRY_VERIFY(!readySpy.isEmpty());
QTRY_VERIFY(!bufferChangedSpy.isEmpty());
QVERIFY(d.bufferAvailable());
QTRY_VERIFY(!durationSpy.isEmpty());
QVERIFY(qAbs(d.duration() - 1000) < 20);
buffer = d.read();
QVERIFY(buffer.isValid());
// See if we got the right format
QVERIFY(buffer.format() == format);
// The decoder should still have the same format
QVERIFY(d.audioFormat() == format);
QVERIFY(errorSpy.isEmpty());
d.stop();
QTRY_COMPARE(d.state(), QAudioDecoder::StoppedState);
QVERIFY(!d.bufferAvailable());
QTRY_COMPARE(durationSpy.count(), 2);
QCOMPARE(d.duration(), qint64(-1));
}
// returns the audio level for each channel
QVector<qreal> getBufferLevels(const QAudioBuffer& buffer)
{
QVector<qreal> values;
if (!buffer.format().isValid() || buffer.format().byteOrder() != QAudioFormat::LittleEndian)
return values;
if (buffer.format().codec() != "audio/pcm")
return values;
int channelCount = buffer.format().channelCount();
values.fill(0, channelCount);
qreal peak_value = getPeakValue(buffer.format());
if (qFuzzyCompare(peak_value, qreal(0)))
return values;
switch (buffer.format().sampleType()) {
case QAudioFormat::Unknown:
case QAudioFormat::UnSignedInt:
if (buffer.format().sampleSize() == 32)
values = getBufferLevels(buffer.constData<quint32>(), buffer.frameCount(), channelCount);
if (buffer.format().sampleSize() == 16)
values = getBufferLevels(buffer.constData<quint16>(), buffer.frameCount(), channelCount);
if (buffer.format().sampleSize() == 8)
values = getBufferLevels(buffer.constData<quint8>(), buffer.frameCount(), channelCount);
for (int i = 0; i < values.size(); ++i)
values[i] = qAbs(values.at(i) - peak_value / 2) / (peak_value / 2);
break;
case QAudioFormat::Float:
if (buffer.format().sampleSize() == 32) {
values = getBufferLevels(buffer.constData<float>(), buffer.frameCount(), channelCount);
for (int i = 0; i < values.size(); ++i)
values[i] /= peak_value;
}
break;
case QAudioFormat::SignedInt:
if (buffer.format().sampleSize() == 32)
values = getBufferLevels(buffer.constData<qint32>(), buffer.frameCount(), channelCount);
if (buffer.format().sampleSize() == 16)
values = getBufferLevels(buffer.constData<qint16>(), buffer.frameCount(), channelCount);
if (buffer.format().sampleSize() == 8)
values = getBufferLevels(buffer.constData<qint8>(), buffer.frameCount(), channelCount);
for (int i = 0; i < values.size(); ++i)
values[i] /= peak_value;
break;
}
return values;
}
// process audio buffer for fft calculations
void MainWindow::processBuffer(QAudioBuffer buffer){
qreal peakValue;
int duration;
if(buffer.frameCount() < 512)
return;
// return left and right audio mean levels
levelLeft = levelRight = 0;
// It only knows how to process stereo audio frames
// mono frames = :P
if(buffer.format().channelCount() != 2)
return;
sample.resize(buffer.frameCount());
// audio is signed int
if(buffer.format().sampleType() == QAudioFormat::SignedInt){
QAudioBuffer::S16S *data = buffer.data<QAudioBuffer::S16S>();
// peak value changes according to sample size.
if (buffer.format().sampleSize() == 32)
peakValue=INT_MAX;
else if (buffer.format().sampleSize() == 16)
peakValue=SHRT_MAX;
else
peakValue=CHAR_MAX;
// scale everything to [0,1]
for(int i=0; i<buffer.frameCount(); i++){
// for visualization purposes, we only need one of the
// left/right channels
sample[i] = data[i].left/peakValue;
levelLeft+= abs(data[i].left)/peakValue;
levelRight+= abs(data[i].right)/peakValue;
}
}
// audio is unsigned int
else if(buffer.format().sampleType() == QAudioFormat::UnSignedInt){
QAudioBuffer::S16U *data = buffer.data<QAudioBuffer::S16U>();
if (buffer.format().sampleSize() == 32)
peakValue=UINT_MAX;
else if (buffer.format().sampleSize() == 16)
peakValue=USHRT_MAX;
else
peakValue=UCHAR_MAX;
for(int i=0; i<buffer.frameCount(); i++){
sample[i] = data[i].left/peakValue;
levelLeft+= abs(data[i].left)/peakValue;
levelRight+= abs(data[i].right)/peakValue;
}
}
// audio is float type
else if(buffer.format().sampleType() == QAudioFormat::Float){
QAudioBuffer::S32F *data = buffer.data<QAudioBuffer::S32F>();
peakValue = 1.00003;
for(int i=0; i<buffer.frameCount(); i++){
sample[i] = data[i].left/peakValue;
// test if sample[i] is infinity (it works)
// some tests produced infinity values :p
if(sample[i] != sample[i]){
sample[i] = 0;
}
else{
levelLeft+= abs(data[i].left)/peakValue;
levelRight+= abs(data[i].right)/peakValue;
}
}
}
// if the probe is listening to the audio
// do fft calculations
// when it is done, calculator will tell us
if(probe->isActive()){
duration = buffer.format().durationForBytes(buffer.frameCount())/1000;
//qDebug() << "duracao =" << duration;
calculator->calc(sample, duration);
}
// tells anyone interested about left and right mean levels
emit levels(levelLeft/buffer.frameCount(),levelRight/buffer.frameCount());
}
void AudioBuffer::init(QAudioBuffer &qtbuffer){
qDebug() << "void AudioBuffer::init(...) called";
QAudioFormat audioFormat = qtbuffer.format();
this->hzFreq = audioFormat.sampleRate();
this->durationInMs = qtbuffer.duration()*1000;
QAudioFormat::SampleType sampleType
= audioFormat.sampleType();
int frameCount = qtbuffer.frameCount();
int nChannels = audioFormat.channelCount();
int bytesPerFrame = audioFormat.bytesPerFrame();
int bytesPerValue = bytesPerFrame/nChannels;
void *firstData = qtbuffer.data();
this->bufferSize = frameCount;
this->buffer = QSharedPointer<SharedBuffer>(
new SharedBuffer);
this->buffer->buffer = new int[this->bufferSize];
this->mean = 0;
const int tempReducFactor = 500;
qint64 meanOfSquare = 0;
for(int i=0; i<frameCount; i++){
int currentValue = 0;
for(int j=0; j<nChannels; j++){
int currentPos
= i*bytesPerFrame
+ j*bytesPerValue;
void *valPos = (void *)(((quint8 *)firstData) + currentPos);
if(sampleType == QAudioFormat::SignedInt){
if(bytesPerValue == 1){
quint8 val = *((quint8*)valPos);
currentValue += val;
}else if(bytesPerValue == 2){
quint16 val = *((quint16*)valPos);
currentValue += val;
}else if(bytesPerValue == 4){
quint32 val = *((quint32*)valPos);
currentValue += val;
}else if(bytesPerValue == 8){
quint64 val = *((quint64*)valPos);
currentValue += val;
}
}else if(sampleType == QAudioFormat::UnSignedInt){
if(bytesPerValue == 1){
qint8 val = *((qint8*)valPos);
currentValue += val;
}else if(bytesPerValue == 2){
qint16 val = *((qint16*)valPos);
currentValue += val;
}else if(bytesPerValue == 4){
qint32 val = *((qint32*)valPos);
currentValue += val;
}else if(bytesPerValue == 8){
qint64 val = *((qint64*)valPos);
currentValue += val;
}
}else if(sampleType == QAudioFormat::Float){
qreal val = *((qreal*)valPos);
currentValue += val;
}
}
currentValue /= nChannels;
currentValue /= tempReducFactor;
this->buffer->buffer[i] = currentValue;
this->mean += currentValue;
meanOfSquare += currentValue*currentValue;
if(meanOfSquare < 0){
Q_ASSERT(false);
}
}
this->mean /= frameCount;
meanOfSquare /= frameCount;
qint64 squaredMean =
this->mean
* this->mean;
this->var = meanOfSquare
- squaredMean;
this->var *= tempReducFactor * tempReducFactor;
this->mean *= tempReducFactor;
this->sd = qSqrt(this->var);
qDebug() << "void AudioBuffer::init(...) end";
}