KeyFinderResultWrapper keyDetectionProcess(const AsyncFileObject& object){
KeyFinderResultWrapper result;
result.batchRow = object.batchRow;
// initialise stream and decode file into it.
KeyFinder::AudioData* audio = NULL;
AudioFileDecoder* decoder = NULL;
try{
decoder = AudioFileDecoder::getDecoder();
}catch(KeyFinder::Exception& e){
delete decoder;
result.errorMessage = QString(e.what().c_str());
return result;
}
try{
audio = decoder->decodeFile(object.filePath);
delete decoder;
}catch(KeyFinder::Exception& e){
delete audio;
delete decoder;
result.errorMessage = QString(e.what().c_str());
return result;
}
// make audio stream monaural ahead of downsample to reduce load
audio->reduceToMono();
// downsample if necessary
if(object.prefs.getDFactor() > 1){
Downsampler* ds = Downsampler::getDownsampler(object.prefs.getDFactor(),audio->getFrameRate(),object.prefs.getLastFreq());
try{
audio = ds->downsample(audio,object.prefs.getDFactor());
}catch(KeyFinder::Exception& e){
delete audio;
delete ds;
result.errorMessage = QString(e.what().c_str());
return result;
}
delete ds;
}
KeyFinder::KeyFinder* kf = LibKeyFinderSingleton::getInstance()->getKeyFinder();
result.core = kf->findKey(*audio, object.prefs.core);
delete audio;
return result;
}
KeyFinderResultWrapper keyDetectionProcess(const AsyncFileObject& object) {
KeyFinderResultWrapper result;
result.batchRow = object.batchRow;
AudioFileDecoder* decoder = NULL;
try {
decoder = new AudioFileDecoder(object.filePath, object.prefs.getMaxDuration());
} catch (std::exception& e) {
delete decoder;
result.errorMessage = QString(e.what());
return result;
} catch (...) {
delete decoder;
result.errorMessage = "Unknown exception initialising decoder";
return result;
}
KeyFinder::Workspace workspace;
static KeyFinder::KeyFinder kf;
try {
while (true) {
KeyFinder::AudioData* tempAudio = decoder->decodeNextAudioPacket();
if (tempAudio == NULL) break;
kf.progressiveChromagram(*tempAudio, workspace);
delete tempAudio;
}
delete decoder;
decoder = NULL;
kf.finalChromagram(workspace);
result.fullChromagram = KeyFinder::Chromagram(*workspace.chromagram);
result.core = kf.keyOfChromagram(workspace);
} catch (std::exception& e) {
if (decoder != NULL) delete decoder;
result.errorMessage = QString(e.what());
return result;
} catch (...) {
if (decoder != NULL) {
delete decoder;
result.errorMessage = "Unknown exception while decoding";
} else {
result.errorMessage = "Unknown exception while analysing";
}
return result;
}
return result;
}
KeyFinderResultWrapper keyDetectionProcess(const AsyncFileObject& object) {
KeyFinderResultWrapper result;
result.batchRow = object.batchRow;
AudioFileDecoder* decoder = NULL;
try {
decoder = new AudioFileDecoder(object.filePath, object.prefs.getMaxDuration());
} catch (std::exception& e) {
delete decoder;
result.errorMessage = QString(e.what());
return result;
} catch (...) {
delete decoder;
result.errorMessage = "Unknown exception initialising decoder";
return result;
}
KeyFinder::Workspace workspace;
static KeyFinder::KeyFinder kf;
try {
while (true) {
KeyFinder::AudioData* tempAudio = decoder->decodeNextAudioPacket();
if (tempAudio == NULL) break;
kf.progressiveChromagram(*tempAudio, workspace, object.prefs.core);
delete tempAudio;
}
delete decoder;
decoder = NULL;
kf.finalChromagram(workspace, object.prefs.core);
result.fullChromagram = KeyFinder::Chromagram(*workspace.chromagram);
result.core = kf.keyOfChromagram(workspace, object.prefs.core);
result.oneOctaveChromagram = result.fullChromagram;
result.oneOctaveChromagram.reduceToOneOctave();
} catch (std::exception& e) {
if (decoder != NULL) delete decoder;
result.errorMessage = QString(e.what());
return result;
} catch (...) {
if (decoder != NULL) {
delete decoder;
result.errorMessage = "Unknown exception while decoding";
} else {
result.errorMessage = "Unknown exception while analysing";
}
return result;
}
for (unsigned int i = 0; i < result.core.segments.size(); i++) {
qDebug(
"Chroma vector for segment %d of file %s: [%.0f, %.0f, %.0f, %.0f, %.0f, %.0f, %.0f, %.0f, %.0f, %.0f, %.0f, %.0f]",
i, object.filePath.toUtf8().constData(),
result.core.segments[i].chromaVector[0], result.core.segments[i].chromaVector[1],
result.core.segments[i].chromaVector[2], result.core.segments[i].chromaVector[3],
result.core.segments[i].chromaVector[4], result.core.segments[i].chromaVector[5],
result.core.segments[i].chromaVector[6], result.core.segments[i].chromaVector[7],
result.core.segments[i].chromaVector[8], result.core.segments[i].chromaVector[9],
result.core.segments[i].chromaVector[10], result.core.segments[i].chromaVector[11]
);
}
return result;
}
void KeyFinderWorkerThread::run(){
if(!haveParams){
emit failed("No parameters.");
return;
}
// initialise stream and decode file into it
AudioStream* astrm = NULL;
AudioFileDecoder* dec = AudioFileDecoder::getDecoder(filePath.toUtf8().data());
try{
astrm = dec->decodeFile(filePath.toUtf8().data());
}catch(Exception){
delete astrm;
delete dec;
emit failed("Could not decode file.");
return;
}
delete dec;
emit decoded();
// make audio stream monaural
astrm->reduceToMono();
emit madeMono();
// downsample if necessary
if(prefs.getDFactor() > 1){
Downsampler* ds = Downsampler::getDownsampler(prefs.getDFactor(),astrm->getFrameRate(),prefs.getLastFreq());
try{
astrm = ds->downsample(astrm,prefs.getDFactor());
}catch(Exception){
delete astrm;
delete ds;
emit failed("Downsampler failed.");
return;
}
delete ds;
emit downsampled();
}
// start spectrum analysis
SpectrumAnalyser* sa = NULL;
Chromagram* ch = NULL;
sa = SpectrumAnalyserFactory::getInstance()->getSpectrumAnalyser(astrm->getFrameRate(),prefs);
ch = sa->chromagram(astrm);
delete astrm; // note we don't delete the spectrum analyser; it stays in the centralised factory for reuse.
ch->reduceTuningBins(prefs);
emit producedFullChromagram(*ch);
// reduce chromagram
ch->reduceToOneOctave(prefs);
emit producedOneOctaveChromagram(*ch);
// get energy level across track to weight segments
std::vector<float> loudness(ch->getHops());
for(int h=0; h<ch->getHops(); h++)
for(int b=0; b<ch->getBins(); b++)
loudness[h] += ch->getMagnitude(h,b);
// get harmonic change signal
Segmentation* hcdf = Segmentation::getSegmentation(prefs);
std::vector<double> harmonicChangeSignal = hcdf->getRateOfChange(ch,prefs);
emit producedHarmonicChangeSignal(harmonicChangeSignal);
// get track segmentation
std::vector<int> changes = hcdf->getSegments(harmonicChangeSignal,prefs);
changes.push_back(ch->getHops()); // It used to be getHops()-1. But this doesn't crash. So we like it.
// batch output of keychange locations for Beatles experiment
//for(int i=1; i<changes.size(); i++) // don't want the leading zero
// std::cout << filePath.substr(53) << "\t" << std::fixed << std::setprecision(2) << changes[i]*(prefs.getHopSize()/(44100.0/prefs.getDFactor())) << std::endl;
// end experiment output
// get key estimates for segments
KeyClassifier hc(prefs);
std::vector<int> keys(0);
std::vector<float> keyWeights(24);
for(int i=0; i<(signed)changes.size()-1; i++){
std::vector<double> chroma(ch->getBins());
for(int j=changes[i]; j<changes[i+1]; j++)
for(int k=0; k<ch->getBins(); k++)
chroma[k] += ch->getMagnitude(j,k);
int key = hc.classify(chroma);
for(int j=changes[i]; j<changes[i+1]; j++){
keys.push_back(key);
if(key < 24) // ignore parts that were classified as silent
keyWeights[key] += loudness[j];
}
}
keys.push_back(keys[keys.size()-1]); // put last key on again to match length of track
delete ch;
emit producedKeyEstimates(keys);
// get global key
int mostCommonKey = 24;
float mostCommonKeyWeight = 0.0;
for(int i=0; i<(signed)keyWeights.size(); i++){
if(keyWeights[i] > mostCommonKeyWeight){
mostCommonKeyWeight = keyWeights[i];
mostCommonKey = i;
}
}
emit producedGlobalKeyEstimate(mostCommonKey);
return;
}
