TEST (AudioDataTest, PrependToInitialised) {
KeyFinder::AudioData a;
KeyFinder::AudioData b;
a.setChannels(1);
a.setFrameRate(1);
ASSERT_THROW(a.prepend(b), KeyFinder::Exception);
b.setChannels(2);
b.setFrameRate(1);
ASSERT_THROW(a.prepend(b), KeyFinder::Exception);
b.setChannels(1);
b.setFrameRate(2);
ASSERT_THROW(a.prepend(b), KeyFinder::Exception);
b.setChannels(1);
b.setFrameRate(1);
ASSERT_NO_THROW(a.prepend(b));
a.addToFrameCount(1);
b.addToFrameCount(1);
a.setSampleByFrame(0, 0, 10.0);
b.setSampleByFrame(0, 0, 20.0);
a.prepend(b);
ASSERT_EQ(2, a.getFrameCount());
ASSERT_FLOAT_EQ(20.0, a.getSampleByFrame(0, 0));
ASSERT_FLOAT_EQ(10.0, a.getSampleByFrame(1, 0));
}
TEST (AudioDataTest, SliceFromBack) {
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(1);
KeyFinder::AudioData* b = NULL;
KeyFinder::AudioData* nullPtr = NULL;
ASSERT_THROW(b = a.sliceSamplesFromBack(1), KeyFinder::Exception);
ASSERT_EQ(nullPtr, b);
a.addToFrameCount(10);
ASSERT_THROW(b = a.sliceSamplesFromBack(11), KeyFinder::Exception);
ASSERT_EQ(nullPtr, b);
a.resetIterators();
float v = 0;
while (a.writeIteratorWithinUpperBound()) {
a.setSampleAtWriteIterator(v);
a.advanceWriteIterator();
v += 1.0;
}
ASSERT_NO_THROW(b = a.sliceSamplesFromBack(5));
ASSERT_NE(nullPtr, b);
ASSERT_EQ(5, a.getSampleCount());
ASSERT_EQ(5, b->getSampleCount());
ASSERT_FLOAT_EQ(5.0, b->getSample(0));
ASSERT_FLOAT_EQ(9.0, b->getSample(4));
delete b;
}
TEST (LowPassFilterTest, WorksOnRepetitiveWaves) {
// make two sine waves, but this time, several seconds long
unsigned int samples = frameRate * 5;
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(frameRate);
a.addToSampleCount(samples);
for (unsigned int i = 0; i < samples; i++) {
float sample = 0.0;
sample += sine_wave(i, highFrequency, frameRate, magnitude); // high freq
sample += sine_wave(i, lowFrequency, frameRate, magnitude); // low freq
a.setSample(i, sample);
// ensure repetition of sine waves is perfect...
if (i >= frameRate) {
ASSERT_NEAR(a.getSample(i), a.getSample(i - frameRate), tolerance);
}
}
KeyFinder::LowPassFilter* lpf = new KeyFinder::LowPassFilter(filterOrder, frameRate, cornerFrequency, filterFFT);
KeyFinder::Workspace w;
lpf->filter(a, w);
delete lpf;
// test for lower wave only
for (unsigned int i = 0; i < samples; i++) {
float expected = sine_wave(i, lowFrequency, frameRate, magnitude);
ASSERT_NEAR(expected, a.getSample(i), tolerance);
}
}
TEST (AudioDataTest, DownsamplerInsistsOnMonophonicAudio) {
KeyFinder::AudioData a;
a.setChannels(2);
a.setFrameRate(100);
a.addToSampleCount(10);
ASSERT_THROW(a.downsample(5), KeyFinder::Exception);
a.reduceToMono();
ASSERT_NO_THROW(a.downsample(5));
}
TEST (LowPassFilterTest, InsistsOnMonophonicAudio) {
KeyFinder::AudioData a;
a.setChannels(2);
a.setFrameRate(frameRate);
a.addToSampleCount(frameRate);
KeyFinder::LowPassFilter* lpf = new KeyFinder::LowPassFilter(filterOrder, frameRate, cornerFrequency, filterFFT);
KeyFinder::Workspace w;
ASSERT_THROW(lpf->filter(a, w), KeyFinder::Exception);
a.reduceToMono();
ASSERT_NO_THROW(lpf->filter(a, w));
delete lpf;
}
TEST (LowPassFilterTest, InitialisesNullBuffer) {
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(frameRate);
a.addToSampleCount(frameRate);
KeyFinder::LowPassFilter* lpf = new KeyFinder::LowPassFilter(filterOrder, frameRate, cornerFrequency, filterFFT);
KeyFinder::Workspace w;
std::vector<double>* nullPtr = NULL;
ASSERT_EQ(nullPtr, w.lpfBuffer);
lpf->filter(a, w);
ASSERT_NE(nullPtr, w.lpfBuffer);
}
TEST (LowPassFilterTest, DoesntAlterAudioMetadata) {
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(frameRate);
a.addToSampleCount(frameRate);
KeyFinder::LowPassFilter* lpf = new KeyFinder::LowPassFilter(filterOrder, frameRate, cornerFrequency, filterFFT);
KeyFinder::Workspace w;
lpf->filter(a, w);
delete lpf;
ASSERT_EQ(1, a.getChannels());
ASSERT_EQ(frameRate, a.getFrameRate());
ASSERT_EQ(frameRate, a.getSampleCount());
}
TEST (AudioDataTest, DiscardFromFront) {
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(1);
ASSERT_THROW(a.discardFramesFromFront(1), KeyFinder::Exception);
a.addToFrameCount(10);
ASSERT_THROW(a.discardFramesFromFront(11), KeyFinder::Exception);
ASSERT_NO_THROW(a.discardFramesFromFront(0));
a.setSampleByFrame(5, 0, 10.0);
ASSERT_NO_THROW(a.discardFramesFromFront(5));
ASSERT_EQ(5, a.getFrameCount());
ASSERT_FLOAT_EQ(10.0, a.getSampleByFrame(0, 0));
}
TEST (AudioDataTest, PrependToNew) {
KeyFinder::AudioData a;
KeyFinder::AudioData b;
ASSERT_NO_THROW(a.prepend(b));
ASSERT_EQ(0, a.getChannels());
ASSERT_EQ(0, a.getFrameRate());
b.setChannels(1);
b.setFrameRate(1);
b.addToFrameCount(1);
ASSERT_NO_THROW(a.prepend(b));
ASSERT_EQ(1, a.getChannels());
ASSERT_EQ(1, a.getFrameRate());
ASSERT_EQ(1, a.getFrameCount());
}
TEST (AudioDataTest, DownsamplerResamplesIntegralRelationship) {
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(100);
a.addToSampleCount(10);
for (unsigned int i = 0; i < 5; i++)
a.setSample(i, 100.0);
for (unsigned int i = 5; i < 10; i++)
a.setSample(i, 500.0);
a.downsample(5);
ASSERT_EQ(20, a.getFrameRate());
ASSERT_EQ(2, a.getSampleCount());
ASSERT_FLOAT_EQ(100.0, a.getSample(0));
ASSERT_FLOAT_EQ(500.0, a.getSample(1));
}
TEST (AudioDataTest, DownsamplerResamplesNonintegralRelationship) {
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(100);
a.addToSampleCount(12);
for (unsigned int i = 0; i < 5; i++)
a.setSample(i, 100.0);
for (unsigned int i = 5; i < 10; i++)
a.setSample(i, 500.0);
for (unsigned int i = 10; i < 12; i++)
a.setSample(i, 1000.0);
a.downsample(5);
ASSERT_EQ(3, a.getSampleCount());
ASSERT_FLOAT_EQ(100.0, a.getSample(0));
ASSERT_FLOAT_EQ(500.0, a.getSample(1));
// this doesn't make total mathematical sense but I'm taking a shortcut for performance
ASSERT_FLOAT_EQ(1000.0, a.getSample(2));
}
TEST (LowPassFilterTest, KillsHigherFreqs) {
// make a high frequency sine wave, one second long
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(frameRate);
a.addToSampleCount(frameRate);
for (unsigned int i = 0; i < frameRate; i++) {
a.setSample(i, sine_wave(i, highFrequency, frameRate, magnitude));
}
KeyFinder::LowPassFilter* lpf = new KeyFinder::LowPassFilter(filterOrder, frameRate, cornerFrequency, filterFFT);
KeyFinder::Workspace w;
lpf->filter(a, w);
delete lpf;
// test for near silence
for (unsigned int i = 0; i < frameRate; i++) {
ASSERT_NEAR(0.0, a.getSample(i), tolerance);
}
}
TEST (LowPassFilterTest, MaintainsLowerFreqs) {
// make a low frequency sine wave, one second long
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(frameRate);
a.addToSampleCount(frameRate);
for (unsigned int i = 0; i < frameRate; i++) {
a.setSample(i, sine_wave(i, lowFrequency, frameRate, magnitude));
}
KeyFinder::LowPassFilter* lpf = new KeyFinder::LowPassFilter(filterOrder, frameRate, cornerFrequency, filterFFT);
KeyFinder::Workspace w;
lpf->filter(a, w);
delete lpf;
// test for near perfect reproduction
for (unsigned int i = 0; i < frameRate; i++) {
float expected = sine_wave(i, lowFrequency, frameRate, magnitude);
ASSERT_NEAR(expected, a.getSample(i), tolerance);
}
}
TEST (AudioDataTest, Iterators) {
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(1);
a.addToSampleCount(10);
a.setSample(0, 10.0);
a.setSample(1, 20.0);
a.setSample(3, 50.0);
a.resetIterators(); // this is required before each use
ASSERT_FLOAT_EQ(10.0, a.getSampleAtReadIterator());
a.setSampleAtWriteIterator(15.0);
ASSERT_FLOAT_EQ(15.0, a.getSampleAtReadIterator());
a.advanceReadIterator();
a.advanceWriteIterator();
ASSERT_FLOAT_EQ(20.0, a.getSampleAtReadIterator());
a.setSampleAtWriteIterator(25.0);
ASSERT_FLOAT_EQ(25.0, a.getSampleAtReadIterator());
a.advanceReadIterator(2);
a.advanceWriteIterator(2);
ASSERT_FLOAT_EQ(50.0, a.getSampleAtReadIterator());
a.setSampleAtWriteIterator(55.0);
ASSERT_FLOAT_EQ(55.0, a.getSampleAtReadIterator());
a.resetIterators();
ASSERT_FLOAT_EQ(15.0, a.getSampleAtReadIterator());
a.setSampleAtWriteIterator(150.0);
ASSERT_FLOAT_EQ(150.0, a.getSampleAtReadIterator());
ASSERT_TRUE(a.readIteratorWithinUpperBound());
ASSERT_TRUE(a.writeIteratorWithinUpperBound());
a.advanceReadIterator(10);
a.advanceWriteIterator(10);
ASSERT_FALSE(a.readIteratorWithinUpperBound());
ASSERT_FALSE(a.writeIteratorWithinUpperBound());
}
TEST (AudioDataTest, DownsamplerResamplesSineWave) {
unsigned int frameRate = 10000;
unsigned int frames = frameRate * 4;
float freq = 20;
float magnitude = 32768.0;
unsigned int factor = 5;
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(frameRate);
a.addToSampleCount(frames);
for (unsigned int i = 0; i < frames; i++)
a.setSample(i, sine_wave(i, freq, frameRate, magnitude));
a.downsample(factor);
unsigned int newFrameRate = frameRate / factor;
unsigned int newFrames = frames / factor;
ASSERT_EQ(newFrameRate, a.getFrameRate());
ASSERT_EQ(newFrames, a.getSampleCount());
for (unsigned int i = 0; i < newFrames; i++)
ASSERT_NEAR(sine_wave(i, freq, newFrameRate, magnitude), a.getSample(i), magnitude * 0.05);
}
TEST (LowPassFilterTest, DoesBothAtOnce) {
// make two sine waves, one second long
KeyFinder::AudioData a;
a.setChannels(1);
a.setFrameRate(frameRate);
a.addToSampleCount(frameRate);
for (unsigned int i = 0; i < frameRate; i++) {
float sample = 0.0;
sample += sine_wave(i, highFrequency, frameRate, magnitude); // high freq
sample += sine_wave(i, lowFrequency, frameRate, magnitude); // low freq
a.setSample(i, sample);
}
KeyFinder::LowPassFilter* lpf = new KeyFinder::LowPassFilter(filterOrder, frameRate, cornerFrequency, filterFFT);
KeyFinder::Workspace w;
lpf->filter(a, w);
delete lpf;
// test for lower wave only
for (unsigned int i = 0; i < frameRate; i++) {
float expected = sine_wave(i, lowFrequency, frameRate, magnitude);
ASSERT_NEAR(expected, a.getSample(i), tolerance);
}
}
KeyFinder::AudioData* LibAvDecoder::decodeFile(const QString& filePath){
QMutexLocker codecMutexLocker(&codecMutex); // mutex the preparatory section of this method
AVCodec *codec = NULL;
AVFormatContext *fCtx = NULL;
AVCodecContext *cCtx = NULL;
AVDictionary* dict = NULL;
// convert filepath
#ifdef Q_OS_WIN
const wchar_t* filePathWc = reinterpret_cast<const wchar_t*>(filePath.constData());
const char* filePathCh = utf16_to_utf8(filePathWc);
#else
QByteArray encodedPath = QFile::encodeName(filePath);
const char* filePathCh = encodedPath;
#endif
// open file
int openInputResult = avformat_open_input(&fCtx, filePathCh, NULL, NULL);
if(openInputResult != 0){
std::ostringstream ss;
ss << "Could not open audio file (" << openInputResult << ")";
throw KeyFinder::Exception(ss.str());
}
if(avformat_find_stream_info(fCtx,NULL) < 0){
throw KeyFinder::Exception("Could not find stream information");
}
int audioStream = -1;
for(int i=0; i<(signed)fCtx->nb_streams; i++){
if(fCtx->streams[i]->codec->codec_type == AVMEDIA_TYPE_AUDIO){
audioStream = i;
break;
}
}
if(audioStream == -1){
throw KeyFinder::Exception("Could not find an audio stream");
}
// Determine stream codec
cCtx = fCtx->streams[audioStream]->codec;
codec = avcodec_find_decoder(cCtx->codec_id);
if(codec == NULL){
throw KeyFinder::Exception("Audio stream has unsupported codec");
}
// Open codec
int codecOpenResult = avcodec_open2(cCtx, codec, &dict);
if(codecOpenResult < 0){
std::ostringstream ss;
ss << "Could not open audio codec: " << codec->long_name << " (" << codecOpenResult << ")";
throw KeyFinder::Exception(ss.str());
}
ReSampleContext* rsCtx = av_audio_resample_init(
cCtx->channels, cCtx->channels,
cCtx->sample_rate, cCtx->sample_rate,
AV_SAMPLE_FMT_S16, cCtx->sample_fmt,
0, 0, 0, 0);
if(rsCtx == NULL){
throw KeyFinder::Exception("Could not create ReSampleContext");
}
qDebug("Decoding %s (%s, %d)", filePathCh, av_get_sample_fmt_name(cCtx->sample_fmt), cCtx->sample_rate);
codecMutexLocker.unlock();
// Prep buffer
KeyFinder::AudioData *audio = new KeyFinder::AudioData();
audio->setFrameRate(cCtx->sample_rate);
audio->setChannels(cCtx->channels);
// Decode stream
AVPacket avpkt;
int badPacketCount = 0;
while(true){
av_init_packet(&avpkt);
if(av_read_frame(fCtx, &avpkt) < 0)
break;
if(avpkt.stream_index == audioStream){
try{
int result = decodePacket(cCtx, rsCtx, &avpkt, audio);
if(result != 0){
if(badPacketCount < 100){
badPacketCount++;
}else{
throw KeyFinder::Exception("100 bad packets");
}
}
}catch(KeyFinder::Exception& e){
throw e;
}
}
av_free_packet(&avpkt);
}
codecMutexLocker.relock();
audio_resample_close(rsCtx);
int codecCloseResult = avcodec_close(cCtx);
if(codecCloseResult < 0){
qCritical("Error closing audio codec: %s (%d)", codec->long_name, codecCloseResult);
}
codecMutexLocker.unlock();
av_close_input_file(fCtx);
return audio;
}
KeyFinder::AudioData* LibAvDecoder::decodeFile(const QString& filePath, const int maxDuration){
QMutexLocker codecMutexLocker(&codecMutex); // mutex the preparatory section of this method
AVCodec *codec = NULL;
AVFormatContext *fCtx = NULL;
AVCodecContext *cCtx = NULL;
AVDictionary* dict = NULL;
// convert filepath
#ifdef Q_OS_WIN
const wchar_t* filePathWc = reinterpret_cast<const wchar_t*>(filePath.constData());
const char* filePathCh = utf16_to_utf8(filePathWc);
#else
QByteArray encodedPath = QFile::encodeName(filePath);
const char* filePathCh = encodedPath;
#endif
// open file
int openInputResult = avformat_open_input(&fCtx, filePathCh, NULL, NULL);
if(openInputResult != 0){
throw KeyFinder::Exception(GuiStrings::getInstance()->libavCouldNotOpenFile(openInputResult).toLocal8Bit().constData());
}
if(avformat_find_stream_info(fCtx, NULL) < 0){
av_close_input_file(fCtx);
throw KeyFinder::Exception(GuiStrings::getInstance()->libavCouldNotFindStreamInformation().toLocal8Bit().constData());
}
int audioStream = -1;
for(int i=0; i<(signed)fCtx->nb_streams; i++){
if(fCtx->streams[i]->codec->codec_type == AVMEDIA_TYPE_AUDIO){
audioStream = i;
break;
}
}
if(audioStream == -1){
av_close_input_file(fCtx);
throw KeyFinder::Exception(GuiStrings::getInstance()->libavCouldNotFindAudioStream().toLocal8Bit().constData());
}
// Determine duration
int durationSeconds = fCtx->duration / AV_TIME_BASE;
int durationMinutes = durationSeconds / 60;
// First condition is a hack for bizarre overestimation of some MP3s
if(durationMinutes < 720 && durationSeconds > maxDuration * 60){
av_close_input_file(fCtx);
throw KeyFinder::Exception(GuiStrings::getInstance()->durationExceedsPreference(durationMinutes, durationSeconds % 60, maxDuration).toLocal8Bit().constData());
}
// Determine stream codec
cCtx = fCtx->streams[audioStream]->codec;
codec = avcodec_find_decoder(cCtx->codec_id);
if(codec == NULL){
av_close_input_file(fCtx);
throw KeyFinder::Exception(GuiStrings::getInstance()->libavUnsupportedCodec().toLocal8Bit().constData());
}
// Open codec
int codecOpenResult = avcodec_open2(cCtx, codec, &dict);
if(codecOpenResult < 0){
av_close_input_file(fCtx);
throw KeyFinder::Exception(GuiStrings::getInstance()->libavCouldNotOpenCodec(codec->long_name, codecOpenResult).toLocal8Bit().constData());
}
ReSampleContext* rsCtx = av_audio_resample_init(
cCtx->channels, cCtx->channels,
cCtx->sample_rate, cCtx->sample_rate,
AV_SAMPLE_FMT_S16, cCtx->sample_fmt,
0, 0, 0, 0);
if(rsCtx == NULL){
avcodec_close(cCtx);
av_close_input_file(fCtx);
throw KeyFinder::Exception(GuiStrings::getInstance()->libavCouldNotCreateResampleContext().toLocal8Bit().constData());
}
qDebug("Decoding %s (%s, %d)", filePathCh, av_get_sample_fmt_name(cCtx->sample_fmt), cCtx->sample_rate);
codecMutexLocker.unlock();
// Prep buffer
KeyFinder::AudioData *audio = new KeyFinder::AudioData();
audio->setFrameRate(cCtx->sample_rate);
audio->setChannels(cCtx->channels);
// Decode stream
AVPacket avpkt;
int badPacketCount = 0;
int badPacketThreshold = 100;
while(true){
av_init_packet(&avpkt);
if(av_read_frame(fCtx, &avpkt) < 0)
break;
if(avpkt.stream_index == audioStream){
try{
int result = decodePacket(cCtx, rsCtx, &avpkt, audio);
if(result != 0){
if(badPacketCount < badPacketThreshold){
badPacketCount++;
}else{
avcodec_close(cCtx);
av_close_input_file(fCtx);
throw KeyFinder::Exception(GuiStrings::getInstance()->libavTooManyBadPackets(badPacketThreshold).toLocal8Bit().constData());
}
}
}catch(KeyFinder::Exception& e){
throw e;
}
}
av_free_packet(&avpkt);
}
codecMutexLocker.relock();
audio_resample_close(rsCtx);
int codecCloseResult = avcodec_close(cCtx);
if(codecCloseResult < 0){
qCritical("Error closing audio codec: %s (%d)", codec->long_name, codecCloseResult);
}
codecMutexLocker.unlock();
av_close_input_file(fCtx);
return audio;
}
TEST (AudioDataTest, FrameRate) {
KeyFinder::AudioData a;
a.setFrameRate(44100);
ASSERT_EQ(44100, a.getFrameRate());
ASSERT_THROW(a.setFrameRate(0), KeyFinder::Exception);
}
const char* kfinder_get_key(short signed int *samples,
unsigned int nb_samples,
short unsigned int frame_rate,
short unsigned int nb_channels)
{
// Check input parameter.
if ((samples == NULL) || (nb_samples == 0) || (frame_rate == 0) || (nb_channels == 0))
{
return "";
}
// Build the main computing object.
KeyFinder::KeyFinder k;
// Build an empty audio object
KeyFinder::AudioData a;
// Prepare the object for your audio stream
a.setFrameRate(frame_rate);
a.setChannels(nb_channels);
a.addToSampleCount(nb_samples);
// Copy your audio into the object (as float).
for (unsigned int i = 0; i < nb_samples; i++)
{
a.setSample(i, (float)samples[i]);
}
// Run the analysis
KeyFinder::key_t r;
try
{
r = k.keyOfAudio(a);
}
catch(const std::exception& e)
{
cerr << "libKeyFinder: exception: " << e.what() << endl;
return "";
}
catch(...)
{
cerr << "libKeyFinder: unknown exception" << endl;
return "";
}
// And do something with the result!
switch(r)
{
case KeyFinder::A_MAJOR: return "AM";
case KeyFinder::A_MINOR: return "Am";
case KeyFinder::B_FLAT_MAJOR: return "BbM";
case KeyFinder::B_FLAT_MINOR: return "Bbm";
case KeyFinder::B_MAJOR: return "BM";
case KeyFinder::B_MINOR: return "Bm";
case KeyFinder::C_MAJOR: return "CM";
case KeyFinder::C_MINOR: return "Cm";
case KeyFinder::D_FLAT_MAJOR: return "DbM";
case KeyFinder::D_FLAT_MINOR: return "Dbm";
case KeyFinder::D_MAJOR: return "DM";
case KeyFinder::D_MINOR: return "Dm";
case KeyFinder::E_FLAT_MAJOR: return "EbM";
case KeyFinder::E_FLAT_MINOR: return "Ebm";
case KeyFinder::E_MAJOR: return "EM";
case KeyFinder::E_MINOR: return "Em";
case KeyFinder::F_MAJOR: return "FM";
case KeyFinder::F_MINOR: return "Fm";
case KeyFinder::G_FLAT_MAJOR: return "GbM";
case KeyFinder::G_FLAT_MINOR: return "Gbm";
case KeyFinder::G_MAJOR: return "GM";
case KeyFinder::G_MINOR: return "Gm";
case KeyFinder::A_FLAT_MAJOR: return "AbM";
case KeyFinder::A_FLAT_MINOR: return "Abm";
case KeyFinder::SILENCE: return "";
default: return "";
}
}